client_golang/prometheus/registry.go

763 lines
26 KiB
Go
Raw Normal View History

// Copyright 2014 The Prometheus Authors
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package prometheus
2012-05-20 01:59:25 +04:00
import (
"bytes"
"errors"
"fmt"
"os"
"sort"
2012-05-20 01:59:25 +04:00
"sync"
2017-08-25 15:50:43 +03:00
"unicode/utf8"
2015-02-27 18:12:59 +03:00
"github.com/golang/protobuf/proto"
dto "github.com/prometheus/client_model/go"
2012-05-20 01:59:25 +04:00
)
const (
// Capacity for the channel to collect metrics and descriptors.
capMetricChan = 1000
capDescChan = 10
)
2012-05-20 01:59:25 +04:00
// DefaultRegisterer and DefaultGatherer are the implementations of the
// Registerer and Gatherer interface a number of convenience functions in this
// package act on. Initially, both variables point to the same Registry, which
// has a process collector (see NewProcessCollector) and a Go collector (see
// NewGoCollector) already registered. This approach to keep default instances
// as global state mirrors the approach of other packages in the Go standard
// library. Note that there are caveats. Change the variables with caution and
// only if you understand the consequences. Users who want to avoid global state
// altogether should not use the convenience function and act on custom
// instances instead.
var (
defaultRegistry = NewRegistry()
DefaultRegisterer Registerer = defaultRegistry
DefaultGatherer Gatherer = defaultRegistry
)
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
func init() {
MustRegister(NewProcessCollector(os.Getpid(), ""))
MustRegister(NewGoCollector())
}
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// NewRegistry creates a new vanilla Registry without any Collectors
// pre-registered.
func NewRegistry() *Registry {
return &Registry{
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
collectorsByID: map[uint64]Collector{},
descIDs: map[uint64]struct{}{},
dimHashesByName: map[string]uint64{},
}
2012-05-20 01:59:25 +04:00
}
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// NewPedanticRegistry returns a registry that checks during collection if each
// collected Metric is consistent with its reported Desc, and if the Desc has
// actually been registered with the registry.
//
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// Usually, a Registry will be happy as long as the union of all collected
// Metrics is consistent and valid even if some metrics are not consistent with
// their own Desc or a Desc provided by their registered Collector. Well-behaved
// Collectors and Metrics will only provide consistent Descs. This Registry is
// useful to test the implementation of Collectors and Metrics.
func NewPedanticRegistry() *Registry {
r := NewRegistry()
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
r.pedanticChecksEnabled = true
return r
}
// Registerer is the interface for the part of a registry in charge of
// registering and unregistering. Users of custom registries should use
// Registerer as type for registration purposes (rather than the Registry type
// directly). In that way, they are free to use custom Registerer implementation
// (e.g. for testing purposes).
type Registerer interface {
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// Register registers a new Collector to be included in metrics
// collection. It returns an error if the descriptors provided by the
// Collector are invalid or if they — in combination with descriptors of
// already registered Collectors — do not fulfill the consistency and
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// uniqueness criteria described in the documentation of metric.Desc.
//
// If the provided Collector is equal to a Collector already registered
// (which includes the case of re-registering the same Collector), the
// returned error is an instance of AlreadyRegisteredError, which
// contains the previously registered Collector.
//
// It is in general not safe to register the same Collector multiple
// times concurrently.
Register(Collector) error
// MustRegister works like Register but registers any number of
// Collectors and panics upon the first registration that causes an
// error.
MustRegister(...Collector)
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// Unregister unregisters the Collector that equals the Collector passed
// in as an argument. (Two Collectors are considered equal if their
// Describe method yields the same set of descriptors.) The function
// returns whether a Collector was unregistered.
//
// Note that even after unregistering, it will not be possible to
// register a new Collector that is inconsistent with the unregistered
// Collector, e.g. a Collector collecting metrics with the same name but
// a different help string. The rationale here is that the same registry
// instance must only collect consistent metrics throughout its
// lifetime.
Unregister(Collector) bool
}
2016-08-04 16:26:27 +03:00
// Gatherer is the interface for the part of a registry in charge of gathering
// the collected metrics into a number of MetricFamilies. The Gatherer interface
// comes with the same general implication as described for the Registerer
// interface.
type Gatherer interface {
// Gather calls the Collect method of the registered Collectors and then
// gathers the collected metrics into a lexicographically sorted slice
// of MetricFamily protobufs. Even if an error occurs, Gather attempts
// to gather as many metrics as possible. Hence, if a non-nil error is
// returned, the returned MetricFamily slice could be nil (in case of a
// fatal error that prevented any meaningful metric collection) or
// contain a number of MetricFamily protobufs, some of which might be
// incomplete, and some might be missing altogether. The returned error
// (which might be a MultiError) explains the details. In scenarios
// where complete collection is critical, the returned MetricFamily
// protobufs should be disregarded if the returned error is non-nil.
Gather() ([]*dto.MetricFamily, error)
}
// Register registers the provided Collector with the DefaultRegisterer.
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
//
// Register is a shortcut for DefaultRegisterer.Register(c). See there for more
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// details.
func Register(c Collector) error {
return DefaultRegisterer.Register(c)
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
}
// MustRegister registers the provided Collectors with the DefaultRegisterer and
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// panics if any error occurs.
//
// MustRegister is a shortcut for DefaultRegisterer.MustRegister(cs...). See
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// there for more details.
func MustRegister(cs ...Collector) {
DefaultRegisterer.MustRegister(cs...)
}
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// Unregister removes the registration of the provided Collector from the
// DefaultRegisterer.
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
//
// Unregister is a shortcut for DefaultRegisterer.Unregister(c). See there for
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// more details.
func Unregister(c Collector) bool {
return DefaultRegisterer.Unregister(c)
}
// GathererFunc turns a function into a Gatherer.
type GathererFunc func() ([]*dto.MetricFamily, error)
// Gather implements Gatherer.
func (gf GathererFunc) Gather() ([]*dto.MetricFamily, error) {
return gf()
}
// AlreadyRegisteredError is returned by the Register method if the Collector to
// be registered has already been registered before, or a different Collector
// that collects the same metrics has been registered before. Registration fails
// in that case, but you can detect from the kind of error what has
// happened. The error contains fields for the existing Collector and the
// (rejected) new Collector that equals the existing one. This can be used to
// find out if an equal Collector has been registered before and switch over to
// using the old one, as demonstrated in the example.
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
type AlreadyRegisteredError struct {
ExistingCollector, NewCollector Collector
2012-05-20 01:59:25 +04:00
}
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
func (err AlreadyRegisteredError) Error() string {
return "duplicate metrics collector registration attempted"
}
// MultiError is a slice of errors implementing the error interface. It is used
// by a Gatherer to report multiple errors during MetricFamily gathering.
type MultiError []error
func (errs MultiError) Error() string {
if len(errs) == 0 {
return ""
}
buf := &bytes.Buffer{}
fmt.Fprintf(buf, "%d error(s) occurred:", len(errs))
for _, err := range errs {
fmt.Fprintf(buf, "\n* %s", err)
}
return buf.String()
}
// MaybeUnwrap returns nil if len(errs) is 0. It returns the first and only
// contained error as error if len(errs is 1). In all other cases, it returns
// the MultiError directly. This is helpful for returning a MultiError in a way
// that only uses the MultiError if needed.
func (errs MultiError) MaybeUnwrap() error {
switch len(errs) {
case 0:
return nil
case 1:
return errs[0]
default:
return errs
}
}
2016-08-04 16:26:27 +03:00
// Registry registers Prometheus collectors, collects their metrics, and gathers
// them into MetricFamilies for exposition. It implements both Registerer and
2016-08-04 16:26:27 +03:00
// Gatherer. The zero value is not usable. Create instances with NewRegistry or
// NewPedanticRegistry.
type Registry struct {
mtx sync.RWMutex
collectorsByID map[uint64]Collector // ID is a hash of the descIDs.
descIDs map[uint64]struct{}
dimHashesByName map[string]uint64
pedanticChecksEnabled bool
}
// Register implements Registerer.
func (r *Registry) Register(c Collector) error {
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
var (
descChan = make(chan *Desc, capDescChan)
newDescIDs = map[uint64]struct{}{}
newDimHashesByName = map[string]uint64{}
collectorID uint64 // Just a sum of all desc IDs.
duplicateDescErr error
)
go func() {
c.Describe(descChan)
close(descChan)
}()
r.mtx.Lock()
defer r.mtx.Unlock()
2017-02-23 12:28:41 +03:00
// Conduct various tests...
for desc := range descChan {
// Is the descriptor valid at all?
if desc.err != nil {
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
return fmt.Errorf("descriptor %s is invalid: %s", desc, desc.err)
}
// Is the descID unique?
// (In other words: Is the fqName + constLabel combination unique?)
if _, exists := r.descIDs[desc.id]; exists {
duplicateDescErr = fmt.Errorf("descriptor %s already exists with the same fully-qualified name and const label values", desc)
}
// If it is not a duplicate desc in this collector, add it to
// the collectorID. (We allow duplicate descs within the same
// collector, but their existence must be a no-op.)
if _, exists := newDescIDs[desc.id]; !exists {
newDescIDs[desc.id] = struct{}{}
collectorID += desc.id
}
// Are all the label names and the help string consistent with
// previous descriptors of the same name?
// First check existing descriptors...
if dimHash, exists := r.dimHashesByName[desc.fqName]; exists {
if dimHash != desc.dimHash {
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
return fmt.Errorf("a previously registered descriptor with the same fully-qualified name as %s has different label names or a different help string", desc)
}
} else {
// ...then check the new descriptors already seen.
if dimHash, exists := newDimHashesByName[desc.fqName]; exists {
if dimHash != desc.dimHash {
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
return fmt.Errorf("descriptors reported by collector have inconsistent label names or help strings for the same fully-qualified name, offender is %s", desc)
}
} else {
newDimHashesByName[desc.fqName] = desc.dimHash
}
}
}
// Did anything happen at all?
if len(newDescIDs) == 0 {
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
return errors.New("collector has no descriptors")
}
if existing, exists := r.collectorsByID[collectorID]; exists {
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
return AlreadyRegisteredError{
ExistingCollector: existing,
NewCollector: c,
}
}
// If the collectorID is new, but at least one of the descs existed
// before, we are in trouble.
if duplicateDescErr != nil {
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
return duplicateDescErr
}
// Only after all tests have passed, actually register.
r.collectorsByID[collectorID] = c
for hash := range newDescIDs {
r.descIDs[hash] = struct{}{}
}
for name, dimHash := range newDimHashesByName {
r.dimHashesByName[name] = dimHash
}
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
return nil
}
// Unregister implements Registerer.
func (r *Registry) Unregister(c Collector) bool {
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
var (
descChan = make(chan *Desc, capDescChan)
descIDs = map[uint64]struct{}{}
collectorID uint64 // Just a sum of the desc IDs.
)
go func() {
c.Describe(descChan)
close(descChan)
}()
for desc := range descChan {
if _, exists := descIDs[desc.id]; !exists {
collectorID += desc.id
descIDs[desc.id] = struct{}{}
}
}
r.mtx.RLock()
if _, exists := r.collectorsByID[collectorID]; !exists {
r.mtx.RUnlock()
return false
}
r.mtx.RUnlock()
r.mtx.Lock()
defer r.mtx.Unlock()
delete(r.collectorsByID, collectorID)
for id := range descIDs {
delete(r.descIDs, id)
}
// dimHashesByName is left untouched as those must be consistent
// throughout the lifetime of a program.
return true
}
// MustRegister implements Registerer.
func (r *Registry) MustRegister(cs ...Collector) {
for _, c := range cs {
if err := r.Register(c); err != nil {
panic(err)
}
}
}
2016-08-04 16:26:27 +03:00
// Gather implements Gatherer.
func (r *Registry) Gather() ([]*dto.MetricFamily, error) {
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
var (
metricChan = make(chan Metric, capMetricChan)
metricHashes = map[uint64]struct{}{}
dimHashes = map[string]uint64{}
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
wg sync.WaitGroup
errs MultiError // The collected errors to return in the end.
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
registeredDescIDs map[uint64]struct{} // Only used for pedantic checks
)
r.mtx.RLock()
metricFamiliesByName := make(map[string]*dto.MetricFamily, len(r.dimHashesByName))
// Scatter.
// (Collectors could be complex and slow, so we call them all at once.)
wg.Add(len(r.collectorsByID))
go func() {
wg.Wait()
close(metricChan)
}()
for _, collector := range r.collectorsByID {
go func(collector Collector) {
defer wg.Done()
collector.Collect(metricChan)
}(collector)
}
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// In case pedantic checks are enabled, we have to copy the map before
// giving up the RLock.
if r.pedanticChecksEnabled {
registeredDescIDs = make(map[uint64]struct{}, len(r.descIDs))
for id := range r.descIDs {
registeredDescIDs[id] = struct{}{}
}
}
r.mtx.RUnlock()
// Drain metricChan in case of premature return.
defer func() {
for range metricChan {
}
}()
// Gather.
for metric := range metricChan {
// This could be done concurrently, too, but it required locking
// of metricFamiliesByName (and of metricHashes if checks are
// enabled). Most likely not worth it.
desc := metric.Desc()
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
dtoMetric := &dto.Metric{}
Allow error reporting during metrics collection and simplify Register(). Both are interface changes I want to get in before public announcement. They only break rare usage cases, and are always easy to fix, but still we want to avoid breaking changes after a wider announcement of the project. The change of Register() simply removes the return of the Collector, which nobody was using in practice. It was just bloating the call syntax. Note that this is different from RegisterOrGet(), which is used at various occasions where you want to register something that might or might not be registered already, but if it is, you want the previously registered Collector back (because that's the relevant one). WRT error reporting: I first tried the obvious way of letting the Collector methods Describe() and Collect() return error. However, I had to conclude that that bloated _many_ calls and their handling in very obnoxious ways. On the other hand, the case where you actually want to report errors during registration or collection is very rare. Hence, this approach has the wrong trade-off. The approach taken here might at first appear clunky but is in practice quite handy, mostly because there is almost no change for the "normal" case of "no special error handling", but also because it plays well with the way descriptors and metrics are handled (via channels). Explaining the approach in more detail: - During registration / describe: Error handling was actually already in place (for invalid descriptors, which carry an error anyway). I only added a convenience function to create an invalid descriptor with a given error on purpose. - Metrics are now treated in a similar way. The Write method returns an error now (the only change in interface). An "invalid metric" is provided that can be sent via the channel to signal that that metric could not be collected. It alse transports an error. NON-GOALS OF THIS COMMIT: This is NOT yet the major improvement of the whole registry part, where we want a public Registry interface and plenty of modular configurations (for error handling, various auto-metrics, http instrumentation, testing, ...). However, we can do that whole thing without breaking existing interfaces. For now (which is a significant issue) any error during collection will either cause a 500 HTTP response or a panic (depending on registry config). Later, we definitely want to have a possibility to skip (and only report somehow) non-collectible metrics instead of aborting the whole scrape.
2015-01-12 21:16:09 +03:00
if err := metric.Write(dtoMetric); err != nil {
errs = append(errs, fmt.Errorf(
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
"error collecting metric %v: %s", desc, err,
))
continue
Allow error reporting during metrics collection and simplify Register(). Both are interface changes I want to get in before public announcement. They only break rare usage cases, and are always easy to fix, but still we want to avoid breaking changes after a wider announcement of the project. The change of Register() simply removes the return of the Collector, which nobody was using in practice. It was just bloating the call syntax. Note that this is different from RegisterOrGet(), which is used at various occasions where you want to register something that might or might not be registered already, but if it is, you want the previously registered Collector back (because that's the relevant one). WRT error reporting: I first tried the obvious way of letting the Collector methods Describe() and Collect() return error. However, I had to conclude that that bloated _many_ calls and their handling in very obnoxious ways. On the other hand, the case where you actually want to report errors during registration or collection is very rare. Hence, this approach has the wrong trade-off. The approach taken here might at first appear clunky but is in practice quite handy, mostly because there is almost no change for the "normal" case of "no special error handling", but also because it plays well with the way descriptors and metrics are handled (via channels). Explaining the approach in more detail: - During registration / describe: Error handling was actually already in place (for invalid descriptors, which carry an error anyway). I only added a convenience function to create an invalid descriptor with a given error on purpose. - Metrics are now treated in a similar way. The Write method returns an error now (the only change in interface). An "invalid metric" is provided that can be sent via the channel to signal that that metric could not be collected. It alse transports an error. NON-GOALS OF THIS COMMIT: This is NOT yet the major improvement of the whole registry part, where we want a public Registry interface and plenty of modular configurations (for error handling, various auto-metrics, http instrumentation, testing, ...). However, we can do that whole thing without breaking existing interfaces. For now (which is a significant issue) any error during collection will either cause a 500 HTTP response or a panic (depending on registry config). Later, we definitely want to have a possibility to skip (and only report somehow) non-collectible metrics instead of aborting the whole scrape.
2015-01-12 21:16:09 +03:00
}
metricFamily, ok := metricFamiliesByName[desc.fqName]
if ok {
if metricFamily.GetHelp() != desc.help {
errs = append(errs, fmt.Errorf(
"collected metric %s %s has help %q but should have %q",
desc.fqName, dtoMetric, desc.help, metricFamily.GetHelp(),
))
continue
}
// TODO(beorn7): Simplify switch once Desc has type.
switch metricFamily.GetType() {
case dto.MetricType_COUNTER:
if dtoMetric.Counter == nil {
errs = append(errs, fmt.Errorf(
"collected metric %s %s should be a Counter",
desc.fqName, dtoMetric,
))
continue
}
case dto.MetricType_GAUGE:
if dtoMetric.Gauge == nil {
errs = append(errs, fmt.Errorf(
"collected metric %s %s should be a Gauge",
desc.fqName, dtoMetric,
))
continue
}
case dto.MetricType_SUMMARY:
if dtoMetric.Summary == nil {
errs = append(errs, fmt.Errorf(
"collected metric %s %s should be a Summary",
desc.fqName, dtoMetric,
))
continue
}
case dto.MetricType_UNTYPED:
if dtoMetric.Untyped == nil {
errs = append(errs, fmt.Errorf(
"collected metric %s %s should be Untyped",
desc.fqName, dtoMetric,
))
continue
}
case dto.MetricType_HISTOGRAM:
if dtoMetric.Histogram == nil {
errs = append(errs, fmt.Errorf(
"collected metric %s %s should be a Histogram",
desc.fqName, dtoMetric,
))
continue
}
default:
panic("encountered MetricFamily with invalid type")
}
} else {
metricFamily = &dto.MetricFamily{}
metricFamily.Name = proto.String(desc.fqName)
metricFamily.Help = proto.String(desc.help)
// TODO(beorn7): Simplify switch once Desc has type.
switch {
case dtoMetric.Gauge != nil:
metricFamily.Type = dto.MetricType_GAUGE.Enum()
case dtoMetric.Counter != nil:
metricFamily.Type = dto.MetricType_COUNTER.Enum()
case dtoMetric.Summary != nil:
metricFamily.Type = dto.MetricType_SUMMARY.Enum()
case dtoMetric.Untyped != nil:
metricFamily.Type = dto.MetricType_UNTYPED.Enum()
case dtoMetric.Histogram != nil:
metricFamily.Type = dto.MetricType_HISTOGRAM.Enum()
default:
errs = append(errs, fmt.Errorf(
"empty metric collected: %s", dtoMetric,
))
continue
}
metricFamiliesByName[desc.fqName] = metricFamily
}
if err := checkMetricConsistency(metricFamily, dtoMetric, metricHashes, dimHashes); err != nil {
errs = append(errs, err)
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
continue
}
if r.pedanticChecksEnabled {
// Is the desc registered at all?
if _, exist := registeredDescIDs[desc.id]; !exist {
errs = append(errs, fmt.Errorf(
"collected metric %s %s with unregistered descriptor %s",
metricFamily.GetName(), dtoMetric, desc,
))
continue
}
if err := checkDescConsistency(metricFamily, dtoMetric, desc); err != nil {
errs = append(errs, err)
continue
}
}
metricFamily.Metric = append(metricFamily.Metric, dtoMetric)
}
return normalizeMetricFamilies(metricFamiliesByName), errs.MaybeUnwrap()
}
// Gatherers is a slice of Gatherer instances that implements the Gatherer
// interface itself. Its Gather method calls Gather on all Gatherers in the
// slice in order and returns the merged results. Errors returned from the
// Gather calles are all returned in a flattened MultiError. Duplicate and
// inconsistent Metrics are skipped (first occurrence in slice order wins) and
// reported in the returned error.
//
// Gatherers can be used to merge the Gather results from multiple
// Registries. It also provides a way to directly inject existing MetricFamily
// protobufs into the gathering by creating a custom Gatherer with a Gather
// method that simply returns the existing MetricFamily protobufs. Note that no
// registration is involved (in contrast to Collector registration), so
// obviously registration-time checks cannot happen. Any inconsistencies between
// the gathered MetricFamilies are reported as errors by the Gather method, and
// inconsistent Metrics are dropped. Invalid parts of the MetricFamilies
// (e.g. syntactically invalid metric or label names) will go undetected.
type Gatherers []Gatherer
// Gather implements Gatherer.
func (gs Gatherers) Gather() ([]*dto.MetricFamily, error) {
var (
metricFamiliesByName = map[string]*dto.MetricFamily{}
metricHashes = map[uint64]struct{}{}
dimHashes = map[string]uint64{}
errs MultiError // The collected errors to return in the end.
)
for i, g := range gs {
mfs, err := g.Gather()
if err != nil {
if multiErr, ok := err.(MultiError); ok {
for _, err := range multiErr {
errs = append(errs, fmt.Errorf("[from Gatherer #%d] %s", i+1, err))
}
} else {
errs = append(errs, fmt.Errorf("[from Gatherer #%d] %s", i+1, err))
}
}
for _, mf := range mfs {
existingMF, exists := metricFamiliesByName[mf.GetName()]
if exists {
if existingMF.GetHelp() != mf.GetHelp() {
errs = append(errs, fmt.Errorf(
"gathered metric family %s has help %q but should have %q",
mf.GetName(), mf.GetHelp(), existingMF.GetHelp(),
))
continue
}
if existingMF.GetType() != mf.GetType() {
errs = append(errs, fmt.Errorf(
"gathered metric family %s has type %s but should have %s",
mf.GetName(), mf.GetType(), existingMF.GetType(),
))
continue
}
} else {
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
existingMF = &dto.MetricFamily{}
existingMF.Name = mf.Name
existingMF.Help = mf.Help
existingMF.Type = mf.Type
metricFamiliesByName[mf.GetName()] = existingMF
}
for _, m := range mf.Metric {
if err := checkMetricConsistency(existingMF, m, metricHashes, dimHashes); err != nil {
errs = append(errs, err)
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
continue
}
existingMF.Metric = append(existingMF.Metric, m)
}
}
}
return normalizeMetricFamilies(metricFamiliesByName), errs.MaybeUnwrap()
}
// metricSorter is a sortable slice of *dto.Metric.
type metricSorter []*dto.Metric
func (s metricSorter) Len() int {
return len(s)
}
func (s metricSorter) Swap(i, j int) {
s[i], s[j] = s[j], s[i]
}
func (s metricSorter) Less(i, j int) bool {
if len(s[i].Label) != len(s[j].Label) {
// This should not happen. The metrics are
// inconsistent. However, we have to deal with the fact, as
// people might use custom collectors or metric family injection
// to create inconsistent metrics. So let's simply compare the
// number of labels in this case. That will still yield
// reproducible sorting.
return len(s[i].Label) < len(s[j].Label)
}
for n, lp := range s[i].Label {
vi := lp.GetValue()
vj := s[j].Label[n].GetValue()
if vi != vj {
return vi < vj
}
}
// We should never arrive here. Multiple metrics with the same
// label set in the same scrape will lead to undefined ingestion
// behavior. However, as above, we have to provide stable sorting
// here, even for inconsistent metrics. So sort equal metrics
// by their timestamp, with missing timestamps (implying "now")
// coming last.
if s[i].TimestampMs == nil {
return false
}
if s[j].TimestampMs == nil {
return true
}
return s[i].GetTimestampMs() < s[j].GetTimestampMs()
}
// normalizeMetricFamilies returns a MetricFamily slice with empty
// MetricFamilies pruned and the remaining MetricFamilies sorted by name within
// the slice, with the contained Metrics sorted within each MetricFamily.
func normalizeMetricFamilies(metricFamiliesByName map[string]*dto.MetricFamily) []*dto.MetricFamily {
for _, mf := range metricFamiliesByName {
sort.Sort(metricSorter(mf.Metric))
}
names := make([]string, 0, len(metricFamiliesByName))
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
for name, mf := range metricFamiliesByName {
if len(mf.Metric) > 0 {
names = append(names, name)
}
}
sort.Strings(names)
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
result := make([]*dto.MetricFamily, 0, len(names))
for _, name := range names {
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
result = append(result, metricFamiliesByName[name])
}
return result
2012-05-20 01:59:25 +04:00
}
// checkMetricConsistency checks if the provided Metric is consistent with the
// provided MetricFamily. It also hashed the Metric labels and the MetricFamily
2017-10-04 19:44:49 +03:00
// name. If the resulting hash is already in the provided metricHashes, an error
// is returned. If not, it is added to metricHashes. The provided dimHashes maps
// MetricFamily names to their dimHash (hashed sorted label names). If dimHashes
// doesn't yet contain a hash for the provided MetricFamily, it is
// added. Otherwise, an error is returned if the existing dimHashes in not equal
// the calculated dimHash.
func checkMetricConsistency(
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
metricFamily *dto.MetricFamily,
dtoMetric *dto.Metric,
metricHashes map[uint64]struct{},
dimHashes map[string]uint64,
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
) error {
// Type consistency with metric family.
if metricFamily.GetType() == dto.MetricType_GAUGE && dtoMetric.Gauge == nil ||
metricFamily.GetType() == dto.MetricType_COUNTER && dtoMetric.Counter == nil ||
metricFamily.GetType() == dto.MetricType_SUMMARY && dtoMetric.Summary == nil ||
metricFamily.GetType() == dto.MetricType_HISTOGRAM && dtoMetric.Histogram == nil ||
metricFamily.GetType() == dto.MetricType_UNTYPED && dtoMetric.Untyped == nil {
return fmt.Errorf(
"collected metric %s %s is not a %s",
metricFamily.GetName(), dtoMetric, metricFamily.GetType(),
)
}
2012-05-20 01:59:25 +04:00
for _, labelPair := range dtoMetric.GetLabel() {
if !utf8.ValidString(*labelPair.Value) {
return fmt.Errorf("collected metric's label %s is not utf8: %#v", *labelPair.Name, *labelPair.Value)
}
}
// Is the metric unique (i.e. no other metric with the same name and the same label values)?
2015-11-12 15:55:12 +03:00
h := hashNew()
h = hashAdd(h, metricFamily.GetName())
h = hashAddByte(h, separatorByte)
dh := hashNew()
// Make sure label pairs are sorted. We depend on it for the consistency
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// check.
sort.Sort(LabelPairSorter(dtoMetric.Label))
for _, lp := range dtoMetric.Label {
2015-11-12 15:55:12 +03:00
h = hashAdd(h, lp.GetValue())
h = hashAddByte(h, separatorByte)
dh = hashAdd(dh, lp.GetName())
dh = hashAddByte(dh, separatorByte)
}
2015-11-12 15:55:12 +03:00
if _, exists := metricHashes[h]; exists {
return fmt.Errorf(
"collected metric %s %s was collected before with the same name and label values",
metricFamily.GetName(), dtoMetric,
)
}
if dimHash, ok := dimHashes[metricFamily.GetName()]; ok {
if dimHash != dh {
return fmt.Errorf(
"collected metric %s %s has label dimensions inconsistent with previously collected metrics in the same metric family",
metricFamily.GetName(), dtoMetric,
)
}
} else {
dimHashes[metricFamily.GetName()] = dh
}
metricHashes[h] = struct{}{}
return nil
}
func checkDescConsistency(
metricFamily *dto.MetricFamily,
dtoMetric *dto.Metric,
desc *Desc,
) error {
Create a public registry interface and separate out HTTP exposition General context and approch =========================== This is the first part of the long awaited wider refurbishment of `client_golang/prometheus/...`. After a lot of struggling, I decided to not go for one breaking big-bang, but cut things into smaller steps after all, mostly to keep the changes manageable and easy to review. I'm aiming for having the invasive breaking changes concentrated in as few steps as possible (ideally one). Some steps will not be breaking at all, but typically there will be breaking changes that only affect quite special cases so that 95+% of users will not be affected. This first step is an example for that, see details below. What's happening in this commit? ================================ This step is about finally creating an exported registry interface. This could not be done by simply export the existing internal implementation because the interface would be _way_ too fat. This commit introduces a qutie lean `Registry` interface (compared to the previous interval implementation). The functions that act on the default registry are retained (with very few exceptions) so that most use cases won't see a change. However, several of those are deprecated now to clean up the namespace in the future. The default registry is kept in the public variable `DefaultRegistry`. This follows the example of the http package in the standard library (cf. `http.DefaultServeMux`, `http.DefaultClient`) with the same implications. (This pattern is somewhat disputed within the Go community but I chose to go with the devil you know instead of creating something more complex or even disallowing any changes to the default registry. The current approach gives everybody the freedom to not touch DefaultRegistry or to do everything with a custom registry to play save.) Another important part in making the registry lean is the extraction of the HTTP exposition, which also allows for customization of the HTTP exposition. Note that the separation of metric collection and exposition has the side effect that managing the MetricFamily and Metric protobuf objects in a free-list or pool isn't really feasible anymore. By now (with better GC in more recent Go versions), the returns were anyway dimisishing. To be effective at all, scrapes had to happen more often than GC cycles, and even then most elements of the protobufs (everything excetp the MetricFamily and Metric structs themselves) would still cause allocation churn. In a future breaking change, the signature of the Write method in the Metric interface will be adjusted accordingly. In this commit, avoiding breakage is more important. The following issues are fixed by this commit (some solved "on the fly" now that I was touching the code anyway and it would have been stupid to port the bugs): https://github.com/prometheus/client_golang/issues/46 https://github.com/prometheus/client_golang/issues/100 https://github.com/prometheus/client_golang/issues/170 https://github.com/prometheus/client_golang/issues/205 Documentation including examples have been amended as required. What future changes does this commit enable? ============================================ The following items are not yet implemented, but this commit opens the possibility of implementing these independently. - The separation of the HTTP exposition allows the implementation of other exposition methods based on the Registry interface, as known from other Prometheus client libraries, e.g. sending the metrics to Graphite. Cf. https://github.com/prometheus/client_golang/issues/197 - The public `Registry` interface allows the implementation of convenience tools for testing metrics collection. Those tools can inspect the collected MetricFamily protobufs and compare them to expectation. Also, tests can use their own testing instance of a registry. Cf. https://github.com/prometheus/client_golang/issues/58 Notable non-goals of this commit ================================ Non-goals that will be tackled later ------------------------------------ The following two issues are quite closely connected to the changes in this commit but the line has been drawn deliberately to address them in later steps of the refurbishment: - `InstrumentHandler` has many known problems. The plan is to create a saner way to conveniently intrument HTTP handlers and remove the old `InstrumentHandler` altogether. To keep breakage low for now, even the default handler to expose metrics is still using the old `InstrumentHandler`. This leads to weird naming inconsistencies but I have deemed it better to not break the world right now but do it in the change that provides better ways of instrumenting HTTP handlers. Cf. https://github.com/prometheus/client_golang/issues/200 - There is work underway to make the whole handling of metric descriptors (`Desc`) more intuitive and transparent for the user (including an ability for less strict checking, cf. https://github.com/prometheus/client_golang/issues/47). That's quite invasive from the perspective of the internal code, namely the registry. I deliberately kept those changes out of this commit. - While this commit adds new external dependency, the effort to vendor anything within the library that is not visible in any exported types will have to be done later. Non-goals that _might_ be tackled later --------------------------------------- There is a strong and understandable urge to divide the `prometheus` package into a number of sub-packages (like `registry`, `collectors`, `http`, `metrics`, …). However, to not run into a multitude of circular import chains, this would need to break every single existing usage of the library. (As just one example, if the ubiquitious `prometheus.MustRegister` (with more than 2,000 uses on GitHub alone) is kept in the `prometheus` package, but the other registry concerns go into a new `registry` package, then the `prometheus` package would import the `registry` package (to call the actual register method), while at the same time the `registry` package needs to import the `prometheus` package to access `Collector`, `Metric`, `Desc` and more. If we moved `MustRegister` into the `registry` package, thousands of code lines would have to be fixed (which would be easy if the world was a mono repo, but it is not). If we moved everything else the proposed registry package needs into packages of their own, we would break thousands of other code lines.) The main problem is really the top-level functions like `MustRegister`, `Handler`, …, which effectively pull everything into one package. Those functions are however very convenient for the easy and very frequent use-cases. This problem has to be revisited later. For now, I'm trying to keep the amount of exported names in the package as low as possible (e.g. I unexported expvarCollector in this commit because the NewExpvarCollector constructor is enough to export, and it is now consistent with other collectors, like the goCollector). Non-goals that won't be tackled anytime soon -------------------------------------------- Something that I have played with a lot is "streaming collection", i.e. allow an implementation of the `Registry` interface that collects metrics incrementally and serves them while doing so. As it has turned out, this has many many issues and makes the `Registry` interface very clunky. Eventually, I made the call that it is unlikely we will really implement streaming collection; and making the interface more clunky for something that might not even happen is really a big no-no. Note that the `Registry` interface only creates the in-memory representation of the metric family protobufs in one go. The serializaton onto the wire can still be handled in a streaming fashion (which hasn't been done so far, without causing any trouble, but might be done in the future without breaking any interfaces). What are the breaking changes? ============================== - Signatures of functions pushing to Pushgateway have changed to allow arbitrary grouping (which was planned for a long time anyway, and now that I had to work on the Push code anyway for the registry refurbishment, I finally did it, cf. https://github.com/prometheus/client_golang/issues/100). With the gained insight that pushing to the default registry is almost never the right thing, and now that we are breaking the Push call anyway, all the Push functions were moved to their own package, which cleans up the namespace and is more idiomatic (pushing Collectors is now literally done by `push.Collectors(...)`). - The registry is doing more consistency checks by default now. Past creators of inconsistent metrics could have masked the problem by not setting `EnableCollectChecks`. Those inconsistencies will now be detected. (But note that a "best effort" metrics collection is now possible with `HandlerOpts.ErrorHandling = ContinueOnError`.) - `EnableCollectChecks` is gone. The registry is now performing some of those checks anyway (see previous item), and a registry with all of those checks can now be created with `NewPedanticRegistry` (only used for testing). - `PanicOnCollectError` is gone. This behavior can now be configured when creating a custom HTTP handler.
2016-07-20 18:11:14 +03:00
// Desc help consistency with metric family help.
if metricFamily.GetHelp() != desc.help {
return fmt.Errorf(
"collected metric %s %s has help %q but should have %q",
metricFamily.GetName(), dtoMetric, metricFamily.GetHelp(), desc.help,
)
}
// Is the desc consistent with the content of the metric?
lpsFromDesc := make([]*dto.LabelPair, 0, len(dtoMetric.Label))
lpsFromDesc = append(lpsFromDesc, desc.constLabelPairs...)
for _, l := range desc.variableLabels {
lpsFromDesc = append(lpsFromDesc, &dto.LabelPair{
Name: proto.String(l),
})
}
if len(lpsFromDesc) != len(dtoMetric.Label) {
return fmt.Errorf(
"labels in collected metric %s %s are inconsistent with descriptor %s",
metricFamily.GetName(), dtoMetric, desc,
)
}
sort.Sort(LabelPairSorter(lpsFromDesc))
for i, lpFromDesc := range lpsFromDesc {
lpFromMetric := dtoMetric.Label[i]
if lpFromDesc.GetName() != lpFromMetric.GetName() ||
lpFromDesc.Value != nil && lpFromDesc.GetValue() != lpFromMetric.GetValue() {
return fmt.Errorf(
"labels in collected metric %s %s are inconsistent with descriptor %s",
metricFamily.GetName(), dtoMetric, desc,
)
}
}
return nil
2012-05-20 01:59:25 +04:00
}