client_golang/prometheus/registry_test.go

1334 lines
36 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.
// Copyright (c) 2013, The Prometheus Authors
// All rights reserved.
//
// Use of this source code is governed by a BSD-style license that can be found
// in the LICENSE file.
package prometheus_test
import (
"bytes"
Added Transactional Gatherer allowed cached solutions (#989) * Added cached collector. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> update. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Attempt 2 Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Added blocking registry, with raw collector and transactional handler. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Added fast path to normal (empty) registry to save 8 allocs and 3K5B per Gather. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Simplified API, added tests. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Fix. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Simplified implementation. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Added benchmark. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Optimized. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Optimization attempt. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Revert "Optimization attempt." This reverts commit 2fcaf51be9a12b4b95413b6b3e0c13fabfaaf73f. Optimization was not worth it: benchstat v1.txt v2.txt name old time/op new time/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 2.64µs ± 0% 4.05µs ± 0% ~ (p=1.000 n=1+1) CachedTGatherer_Update/Update_of_all_elements_with_reset-12 701ms ± 0% 358ms ± 0% ~ (p=1.000 n=1+1) CachedTGatherer_Update/Gather-12 535µs ± 0% 703934µs ± 0% ~ (p=1.000 n=1+1) name old alloc/op new alloc/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 208B ± 0% 208B ± 0% ~ (all equal) CachedTGatherer_Update/Update_of_all_elements_with_reset-12 40.2MB ± 0% 41.1MB ± 0% ~ (p=1.000 n=1+1) CachedTGatherer_Update/Gather-12 48.6kB ± 0% 84.3kB ± 0% ~ (p=1.000 n=1+1) name old allocs/op new allocs/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 3.00 ± 0% 3.00 ± 0% ~ (all equal) CachedTGatherer_Update/Update_of_all_elements_with_reset-12 6.00 ± 0% 4003.00 ± 0% ~ (p=1.000 n=1+1) CachedTGatherer_Update/Gather-12 1.00k ± 0% 2.01k ± 0% ~ (p=1.000 n=1+1) * nit. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Another optimization attempt. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * rename and further optimization. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Hopefully final optimization. benchstat -delta-test=none v6.txt v9.txt name old time/op new time/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 13.1ms ± 0% 0.0ms ± 0% -99.81% CachedTGatherer_Update/Update_of_all_elements_with_reset-12 309ms ± 0% 282ms ± 0% -8.77% CachedTGatherer_Update/Gather-12 422ms ± 0% 0ms ± 0% -99.95% name old alloc/op new alloc/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 208B ± 0% 208B ± 0% 0.00% CachedTGatherer_Update/Update_of_all_elements_with_reset-12 2.47kB ± 0% 1.67kB ± 0% -32.56% CachedTGatherer_Update/Gather-12 52.8kB ± 0% 24.6kB ± 0% -53.34% name old allocs/op new allocs/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 3.00 ± 0% 3.00 ± 0% 0.00% CachedTGatherer_Update/Update_of_all_elements_with_reset-12 0.00 0.00 0.00% CachedTGatherer_Update/Gather-12 1.00k ± 0% 0.00k ± 0% -99.60% Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Removed obsolete comment Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Fixed tests. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Removed cache. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Fixed tests. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Re-add cache. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Removed cache. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com>
2022-02-23 14:22:52 +03:00
"errors"
"fmt"
"math/rand"
"net/http"
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
"net/http/httptest"
"os"
"sync"
"testing"
"time"
"github.com/prometheus/client_golang/prometheus"
"github.com/prometheus/client_golang/prometheus/promhttp"
dto "github.com/prometheus/client_model/go"
"github.com/prometheus/common/expfmt"
"google.golang.org/protobuf/proto"
)
// uncheckedCollector wraps a Collector but its Describe method yields no Desc.
type uncheckedCollector struct {
c prometheus.Collector
}
func (u uncheckedCollector) Describe(_ chan<- *prometheus.Desc) {}
func (u uncheckedCollector) Collect(c chan<- prometheus.Metric) {
u.c.Collect(c)
}
func testHandler(t testing.TB) {
// TODO(beorn7): This test is a bit too "end-to-end". It tests quite a
// few moving parts that are not strongly coupled. They could/should be
// tested separately. However, the changes planned for v2 will
// require a major rework of this test anyway, at which time I will
// structure it in a better way.
metricVec := prometheus.NewCounterVec(
prometheus.CounterOpts{
Name: "name",
Help: "docstring",
ConstLabels: prometheus.Labels{"constname": "constvalue"},
},
[]string{"labelname"},
)
metricVec.WithLabelValues("val1").Inc()
metricVec.WithLabelValues("val2").Inc()
externalMetricFamily := &dto.MetricFamily{
Name: proto.String("externalname"),
Help: proto.String("externaldocstring"),
Type: dto.MetricType_COUNTER.Enum(),
Metric: []*dto.Metric{
{
Label: []*dto.LabelPair{
{
Name: proto.String("externalconstname"),
Value: proto.String("externalconstvalue"),
},
{
Name: proto.String("externallabelname"),
Value: proto.String("externalval1"),
},
},
Counter: &dto.Counter{
Value: proto.Float64(1),
},
},
},
}
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
externalBuf := &bytes.Buffer{}
enc := expfmt.NewEncoder(externalBuf, expfmt.FmtProtoDelim)
if err := enc.Encode(externalMetricFamily); err != nil {
t.Fatal(err)
}
externalMetricFamilyAsBytes := externalBuf.Bytes()
externalMetricFamilyAsText := []byte(`# HELP externalname externaldocstring
# TYPE externalname counter
externalname{externalconstname="externalconstvalue",externallabelname="externalval1"} 1
`)
externalMetricFamilyAsProtoText := []byte(`name: "externalname"
help: "externaldocstring"
type: COUNTER
metric: <
label: <
name: "externalconstname"
value: "externalconstvalue"
>
label: <
name: "externallabelname"
value: "externalval1"
>
counter: <
value: 1
>
>
`)
externalMetricFamilyAsProtoCompactText := []byte(`name:"externalname" help:"externaldocstring" type:COUNTER metric:<label:<name:"externalconstname" value:"externalconstvalue" > label:<name:"externallabelname" value:"externalval1" > counter:<value:1 > >`)
externalMetricFamilyAsProtoCompactText = append(externalMetricFamilyAsProtoCompactText, []byte(" \n")...)
expectedMetricFamily := &dto.MetricFamily{
Name: proto.String("name"),
Help: proto.String("docstring"),
Type: dto.MetricType_COUNTER.Enum(),
Metric: []*dto.Metric{
Optimize fingerprinting and metric locks. These are all simple changes we should have caught a long time ago: 1. The hashing mechanism for fingerprint label sets should have not allocated new objects for the actual hashing---at least not egregiously. This simplifies the hash writing by just byte- dumping the string stream into the hasher. 2. The hashing mechanism within the scope of a metric does not care about the value of the label keys themselves but only of the label values. The keys can be dropped from the calculation. 3. The locking mechanism for the metrics should not block on hash computation but rather solely on the actual mutation or critical section reads. 4. For scalar metrics (i.e., ones with niladic label signatures), we should rely on a preallocated map versus requesting a new one ad hoc. This is tested with Go 1.1, so the results may yield other values for us elsewhere: BEFORE BenchmarkLabelValuesToSignatureScalar 500000000 3.97 ns/op 0 B/op 0 allocs/op BenchmarkLabelValuesToSignatureSingle 5000000 714 ns/op 74 B/op 4 allocs/op BenchmarkLabelValuesToSignatureDouble 1000000 1153 ns/op 107 B/op 5 allocs/op BenchmarkLabelValuesToSignatureTriple 1000000 1588 ns/op 138 B/op 6 allocs/op BenchmarkLabelToSignatureScalar 500000000 3.91 ns/op 0 B/op 0 allocs/op BenchmarkLabelToSignatureSingle 2000000 874 ns/op 92 B/op 5 allocs/op BenchmarkLabelToSignatureDouble 1000000 1528 ns/op 139 B/op 7 allocs/op BenchmarkLabelToSignatureTriple 1000000 2172 ns/op 186 B/op 9 allocs/op AFTER BenchmarkLabelValuesToSignatureScalar 500000000 4.36 ns/op 0 B/op 0 allocs/op BenchmarkLabelValuesToSignatureSingle 5000000 378 ns/op 89 B/op 4 allocs/op BenchmarkLabelValuesToSignatureDouble 5000000 574 ns/op 142 B/op 5 allocs/op BenchmarkLabelValuesToSignatureTriple 5000000 758 ns/op 186 B/op 6 allocs/op BenchmarkLabelToSignatureScalar 500000000 4.06 ns/op 0 B/op 0 allocs/op BenchmarkLabelToSignatureSingle 5000000 472 ns/op 106 B/op 5 allocs/op BenchmarkLabelToSignatureDouble 2000000 746 ns/op 174 B/op 7 allocs/op BenchmarkLabelToSignatureTriple 1000000 1061 ns/op 235 B/op 9 allocs/op In effect, a single metric mutation operation's lookup overhead will move from Before::iBenchmarkLabelToSignature to After::BenchmarkLabelValuesToSignature. This MINIMALLY reduces 1/2 the overhead. I would be hesitant in reading the memory allocation statistics, for this was run with the GC still on and thusly inaccurate per Go benchmarking documentation. Before::BenchmarkLabelValuesToSignature never existed, so it is not of any intrinsic value in itself. That said, the cases that still rely on LabelToSignature experience consistently a 1/2 drop in time. Change-Id: Ifc9e69f718af65a59f5be8117473518233258159
2014-04-14 20:45:16 +04:00
{
Label: []*dto.LabelPair{
Optimize fingerprinting and metric locks. These are all simple changes we should have caught a long time ago: 1. The hashing mechanism for fingerprint label sets should have not allocated new objects for the actual hashing---at least not egregiously. This simplifies the hash writing by just byte- dumping the string stream into the hasher. 2. The hashing mechanism within the scope of a metric does not care about the value of the label keys themselves but only of the label values. The keys can be dropped from the calculation. 3. The locking mechanism for the metrics should not block on hash computation but rather solely on the actual mutation or critical section reads. 4. For scalar metrics (i.e., ones with niladic label signatures), we should rely on a preallocated map versus requesting a new one ad hoc. This is tested with Go 1.1, so the results may yield other values for us elsewhere: BEFORE BenchmarkLabelValuesToSignatureScalar 500000000 3.97 ns/op 0 B/op 0 allocs/op BenchmarkLabelValuesToSignatureSingle 5000000 714 ns/op 74 B/op 4 allocs/op BenchmarkLabelValuesToSignatureDouble 1000000 1153 ns/op 107 B/op 5 allocs/op BenchmarkLabelValuesToSignatureTriple 1000000 1588 ns/op 138 B/op 6 allocs/op BenchmarkLabelToSignatureScalar 500000000 3.91 ns/op 0 B/op 0 allocs/op BenchmarkLabelToSignatureSingle 2000000 874 ns/op 92 B/op 5 allocs/op BenchmarkLabelToSignatureDouble 1000000 1528 ns/op 139 B/op 7 allocs/op BenchmarkLabelToSignatureTriple 1000000 2172 ns/op 186 B/op 9 allocs/op AFTER BenchmarkLabelValuesToSignatureScalar 500000000 4.36 ns/op 0 B/op 0 allocs/op BenchmarkLabelValuesToSignatureSingle 5000000 378 ns/op 89 B/op 4 allocs/op BenchmarkLabelValuesToSignatureDouble 5000000 574 ns/op 142 B/op 5 allocs/op BenchmarkLabelValuesToSignatureTriple 5000000 758 ns/op 186 B/op 6 allocs/op BenchmarkLabelToSignatureScalar 500000000 4.06 ns/op 0 B/op 0 allocs/op BenchmarkLabelToSignatureSingle 5000000 472 ns/op 106 B/op 5 allocs/op BenchmarkLabelToSignatureDouble 2000000 746 ns/op 174 B/op 7 allocs/op BenchmarkLabelToSignatureTriple 1000000 1061 ns/op 235 B/op 9 allocs/op In effect, a single metric mutation operation's lookup overhead will move from Before::iBenchmarkLabelToSignature to After::BenchmarkLabelValuesToSignature. This MINIMALLY reduces 1/2 the overhead. I would be hesitant in reading the memory allocation statistics, for this was run with the GC still on and thusly inaccurate per Go benchmarking documentation. Before::BenchmarkLabelValuesToSignature never existed, so it is not of any intrinsic value in itself. That said, the cases that still rely on LabelToSignature experience consistently a 1/2 drop in time. Change-Id: Ifc9e69f718af65a59f5be8117473518233258159
2014-04-14 20:45:16 +04:00
{
Name: proto.String("constname"),
Value: proto.String("constvalue"),
},
Optimize fingerprinting and metric locks. These are all simple changes we should have caught a long time ago: 1. The hashing mechanism for fingerprint label sets should have not allocated new objects for the actual hashing---at least not egregiously. This simplifies the hash writing by just byte- dumping the string stream into the hasher. 2. The hashing mechanism within the scope of a metric does not care about the value of the label keys themselves but only of the label values. The keys can be dropped from the calculation. 3. The locking mechanism for the metrics should not block on hash computation but rather solely on the actual mutation or critical section reads. 4. For scalar metrics (i.e., ones with niladic label signatures), we should rely on a preallocated map versus requesting a new one ad hoc. This is tested with Go 1.1, so the results may yield other values for us elsewhere: BEFORE BenchmarkLabelValuesToSignatureScalar 500000000 3.97 ns/op 0 B/op 0 allocs/op BenchmarkLabelValuesToSignatureSingle 5000000 714 ns/op 74 B/op 4 allocs/op BenchmarkLabelValuesToSignatureDouble 1000000 1153 ns/op 107 B/op 5 allocs/op BenchmarkLabelValuesToSignatureTriple 1000000 1588 ns/op 138 B/op 6 allocs/op BenchmarkLabelToSignatureScalar 500000000 3.91 ns/op 0 B/op 0 allocs/op BenchmarkLabelToSignatureSingle 2000000 874 ns/op 92 B/op 5 allocs/op BenchmarkLabelToSignatureDouble 1000000 1528 ns/op 139 B/op 7 allocs/op BenchmarkLabelToSignatureTriple 1000000 2172 ns/op 186 B/op 9 allocs/op AFTER BenchmarkLabelValuesToSignatureScalar 500000000 4.36 ns/op 0 B/op 0 allocs/op BenchmarkLabelValuesToSignatureSingle 5000000 378 ns/op 89 B/op 4 allocs/op BenchmarkLabelValuesToSignatureDouble 5000000 574 ns/op 142 B/op 5 allocs/op BenchmarkLabelValuesToSignatureTriple 5000000 758 ns/op 186 B/op 6 allocs/op BenchmarkLabelToSignatureScalar 500000000 4.06 ns/op 0 B/op 0 allocs/op BenchmarkLabelToSignatureSingle 5000000 472 ns/op 106 B/op 5 allocs/op BenchmarkLabelToSignatureDouble 2000000 746 ns/op 174 B/op 7 allocs/op BenchmarkLabelToSignatureTriple 1000000 1061 ns/op 235 B/op 9 allocs/op In effect, a single metric mutation operation's lookup overhead will move from Before::iBenchmarkLabelToSignature to After::BenchmarkLabelValuesToSignature. This MINIMALLY reduces 1/2 the overhead. I would be hesitant in reading the memory allocation statistics, for this was run with the GC still on and thusly inaccurate per Go benchmarking documentation. Before::BenchmarkLabelValuesToSignature never existed, so it is not of any intrinsic value in itself. That said, the cases that still rely on LabelToSignature experience consistently a 1/2 drop in time. Change-Id: Ifc9e69f718af65a59f5be8117473518233258159
2014-04-14 20:45:16 +04:00
{
Name: proto.String("labelname"),
Value: proto.String("val1"),
},
},
Counter: &dto.Counter{
Value: proto.Float64(1),
},
},
Optimize fingerprinting and metric locks. These are all simple changes we should have caught a long time ago: 1. The hashing mechanism for fingerprint label sets should have not allocated new objects for the actual hashing---at least not egregiously. This simplifies the hash writing by just byte- dumping the string stream into the hasher. 2. The hashing mechanism within the scope of a metric does not care about the value of the label keys themselves but only of the label values. The keys can be dropped from the calculation. 3. The locking mechanism for the metrics should not block on hash computation but rather solely on the actual mutation or critical section reads. 4. For scalar metrics (i.e., ones with niladic label signatures), we should rely on a preallocated map versus requesting a new one ad hoc. This is tested with Go 1.1, so the results may yield other values for us elsewhere: BEFORE BenchmarkLabelValuesToSignatureScalar 500000000 3.97 ns/op 0 B/op 0 allocs/op BenchmarkLabelValuesToSignatureSingle 5000000 714 ns/op 74 B/op 4 allocs/op BenchmarkLabelValuesToSignatureDouble 1000000 1153 ns/op 107 B/op 5 allocs/op BenchmarkLabelValuesToSignatureTriple 1000000 1588 ns/op 138 B/op 6 allocs/op BenchmarkLabelToSignatureScalar 500000000 3.91 ns/op 0 B/op 0 allocs/op BenchmarkLabelToSignatureSingle 2000000 874 ns/op 92 B/op 5 allocs/op BenchmarkLabelToSignatureDouble 1000000 1528 ns/op 139 B/op 7 allocs/op BenchmarkLabelToSignatureTriple 1000000 2172 ns/op 186 B/op 9 allocs/op AFTER BenchmarkLabelValuesToSignatureScalar 500000000 4.36 ns/op 0 B/op 0 allocs/op BenchmarkLabelValuesToSignatureSingle 5000000 378 ns/op 89 B/op 4 allocs/op BenchmarkLabelValuesToSignatureDouble 5000000 574 ns/op 142 B/op 5 allocs/op BenchmarkLabelValuesToSignatureTriple 5000000 758 ns/op 186 B/op 6 allocs/op BenchmarkLabelToSignatureScalar 500000000 4.06 ns/op 0 B/op 0 allocs/op BenchmarkLabelToSignatureSingle 5000000 472 ns/op 106 B/op 5 allocs/op BenchmarkLabelToSignatureDouble 2000000 746 ns/op 174 B/op 7 allocs/op BenchmarkLabelToSignatureTriple 1000000 1061 ns/op 235 B/op 9 allocs/op In effect, a single metric mutation operation's lookup overhead will move from Before::iBenchmarkLabelToSignature to After::BenchmarkLabelValuesToSignature. This MINIMALLY reduces 1/2 the overhead. I would be hesitant in reading the memory allocation statistics, for this was run with the GC still on and thusly inaccurate per Go benchmarking documentation. Before::BenchmarkLabelValuesToSignature never existed, so it is not of any intrinsic value in itself. That said, the cases that still rely on LabelToSignature experience consistently a 1/2 drop in time. Change-Id: Ifc9e69f718af65a59f5be8117473518233258159
2014-04-14 20:45:16 +04:00
{
Label: []*dto.LabelPair{
Optimize fingerprinting and metric locks. These are all simple changes we should have caught a long time ago: 1. The hashing mechanism for fingerprint label sets should have not allocated new objects for the actual hashing---at least not egregiously. This simplifies the hash writing by just byte- dumping the string stream into the hasher. 2. The hashing mechanism within the scope of a metric does not care about the value of the label keys themselves but only of the label values. The keys can be dropped from the calculation. 3. The locking mechanism for the metrics should not block on hash computation but rather solely on the actual mutation or critical section reads. 4. For scalar metrics (i.e., ones with niladic label signatures), we should rely on a preallocated map versus requesting a new one ad hoc. This is tested with Go 1.1, so the results may yield other values for us elsewhere: BEFORE BenchmarkLabelValuesToSignatureScalar 500000000 3.97 ns/op 0 B/op 0 allocs/op BenchmarkLabelValuesToSignatureSingle 5000000 714 ns/op 74 B/op 4 allocs/op BenchmarkLabelValuesToSignatureDouble 1000000 1153 ns/op 107 B/op 5 allocs/op BenchmarkLabelValuesToSignatureTriple 1000000 1588 ns/op 138 B/op 6 allocs/op BenchmarkLabelToSignatureScalar 500000000 3.91 ns/op 0 B/op 0 allocs/op BenchmarkLabelToSignatureSingle 2000000 874 ns/op 92 B/op 5 allocs/op BenchmarkLabelToSignatureDouble 1000000 1528 ns/op 139 B/op 7 allocs/op BenchmarkLabelToSignatureTriple 1000000 2172 ns/op 186 B/op 9 allocs/op AFTER BenchmarkLabelValuesToSignatureScalar 500000000 4.36 ns/op 0 B/op 0 allocs/op BenchmarkLabelValuesToSignatureSingle 5000000 378 ns/op 89 B/op 4 allocs/op BenchmarkLabelValuesToSignatureDouble 5000000 574 ns/op 142 B/op 5 allocs/op BenchmarkLabelValuesToSignatureTriple 5000000 758 ns/op 186 B/op 6 allocs/op BenchmarkLabelToSignatureScalar 500000000 4.06 ns/op 0 B/op 0 allocs/op BenchmarkLabelToSignatureSingle 5000000 472 ns/op 106 B/op 5 allocs/op BenchmarkLabelToSignatureDouble 2000000 746 ns/op 174 B/op 7 allocs/op BenchmarkLabelToSignatureTriple 1000000 1061 ns/op 235 B/op 9 allocs/op In effect, a single metric mutation operation's lookup overhead will move from Before::iBenchmarkLabelToSignature to After::BenchmarkLabelValuesToSignature. This MINIMALLY reduces 1/2 the overhead. I would be hesitant in reading the memory allocation statistics, for this was run with the GC still on and thusly inaccurate per Go benchmarking documentation. Before::BenchmarkLabelValuesToSignature never existed, so it is not of any intrinsic value in itself. That said, the cases that still rely on LabelToSignature experience consistently a 1/2 drop in time. Change-Id: Ifc9e69f718af65a59f5be8117473518233258159
2014-04-14 20:45:16 +04:00
{
Name: proto.String("constname"),
Value: proto.String("constvalue"),
},
Optimize fingerprinting and metric locks. These are all simple changes we should have caught a long time ago: 1. The hashing mechanism for fingerprint label sets should have not allocated new objects for the actual hashing---at least not egregiously. This simplifies the hash writing by just byte- dumping the string stream into the hasher. 2. The hashing mechanism within the scope of a metric does not care about the value of the label keys themselves but only of the label values. The keys can be dropped from the calculation. 3. The locking mechanism for the metrics should not block on hash computation but rather solely on the actual mutation or critical section reads. 4. For scalar metrics (i.e., ones with niladic label signatures), we should rely on a preallocated map versus requesting a new one ad hoc. This is tested with Go 1.1, so the results may yield other values for us elsewhere: BEFORE BenchmarkLabelValuesToSignatureScalar 500000000 3.97 ns/op 0 B/op 0 allocs/op BenchmarkLabelValuesToSignatureSingle 5000000 714 ns/op 74 B/op 4 allocs/op BenchmarkLabelValuesToSignatureDouble 1000000 1153 ns/op 107 B/op 5 allocs/op BenchmarkLabelValuesToSignatureTriple 1000000 1588 ns/op 138 B/op 6 allocs/op BenchmarkLabelToSignatureScalar 500000000 3.91 ns/op 0 B/op 0 allocs/op BenchmarkLabelToSignatureSingle 2000000 874 ns/op 92 B/op 5 allocs/op BenchmarkLabelToSignatureDouble 1000000 1528 ns/op 139 B/op 7 allocs/op BenchmarkLabelToSignatureTriple 1000000 2172 ns/op 186 B/op 9 allocs/op AFTER BenchmarkLabelValuesToSignatureScalar 500000000 4.36 ns/op 0 B/op 0 allocs/op BenchmarkLabelValuesToSignatureSingle 5000000 378 ns/op 89 B/op 4 allocs/op BenchmarkLabelValuesToSignatureDouble 5000000 574 ns/op 142 B/op 5 allocs/op BenchmarkLabelValuesToSignatureTriple 5000000 758 ns/op 186 B/op 6 allocs/op BenchmarkLabelToSignatureScalar 500000000 4.06 ns/op 0 B/op 0 allocs/op BenchmarkLabelToSignatureSingle 5000000 472 ns/op 106 B/op 5 allocs/op BenchmarkLabelToSignatureDouble 2000000 746 ns/op 174 B/op 7 allocs/op BenchmarkLabelToSignatureTriple 1000000 1061 ns/op 235 B/op 9 allocs/op In effect, a single metric mutation operation's lookup overhead will move from Before::iBenchmarkLabelToSignature to After::BenchmarkLabelValuesToSignature. This MINIMALLY reduces 1/2 the overhead. I would be hesitant in reading the memory allocation statistics, for this was run with the GC still on and thusly inaccurate per Go benchmarking documentation. Before::BenchmarkLabelValuesToSignature never existed, so it is not of any intrinsic value in itself. That said, the cases that still rely on LabelToSignature experience consistently a 1/2 drop in time. Change-Id: Ifc9e69f718af65a59f5be8117473518233258159
2014-04-14 20:45:16 +04:00
{
Name: proto.String("labelname"),
Value: proto.String("val2"),
},
},
Counter: &dto.Counter{
Value: proto.Float64(1),
},
},
},
}
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
buf := &bytes.Buffer{}
enc = expfmt.NewEncoder(buf, expfmt.FmtProtoDelim)
if err := enc.Encode(expectedMetricFamily); err != nil {
t.Fatal(err)
}
expectedMetricFamilyAsBytes := buf.Bytes()
expectedMetricFamilyAsText := []byte(`# HELP name docstring
# TYPE name counter
name{constname="constvalue",labelname="val1"} 1
name{constname="constvalue",labelname="val2"} 1
`)
expectedMetricFamilyAsProtoText := []byte(`name: "name"
help: "docstring"
type: COUNTER
metric: <
label: <
name: "constname"
value: "constvalue"
>
label: <
name: "labelname"
value: "val1"
>
counter: <
value: 1
>
>
metric: <
label: <
name: "constname"
value: "constvalue"
>
label: <
name: "labelname"
value: "val2"
>
counter: <
value: 1
>
>
`)
expectedMetricFamilyAsProtoCompactText := []byte(`name:"name" help:"docstring" type:COUNTER metric:<label:<name:"constname" value:"constvalue" > label:<name:"labelname" value:"val1" > counter:<value:1 > > metric:<label:<name:"constname" value:"constvalue" > label:<name:"labelname" value:"val2" > counter:<value:1 > >`)
expectedMetricFamilyAsProtoCompactText = append(expectedMetricFamilyAsProtoCompactText, []byte(" \n")...)
externalMetricFamilyWithSameName := &dto.MetricFamily{
Name: proto.String("name"),
Help: proto.String("docstring"),
Type: dto.MetricType_COUNTER.Enum(),
Metric: []*dto.Metric{
{
Label: []*dto.LabelPair{
{
Name: proto.String("constname"),
Value: proto.String("constvalue"),
},
{
Name: proto.String("labelname"),
Value: proto.String("different_val"),
},
},
Counter: &dto.Counter{
Value: proto.Float64(42),
},
},
},
}
expectedMetricFamilyMergedWithExternalAsProtoCompactText := []byte(`name:"name" help:"docstring" type:COUNTER metric:<label:<name:"constname" value:"constvalue" > label:<name:"labelname" value:"different_val" > counter:<value:42 > > metric:<label:<name:"constname" value:"constvalue" > label:<name:"labelname" value:"val1" > counter:<value:1 > > metric:<label:<name:"constname" value:"constvalue" > label:<name:"labelname" value:"val2" > counter:<value:1 > >`)
expectedMetricFamilyMergedWithExternalAsProtoCompactText = append(expectedMetricFamilyMergedWithExternalAsProtoCompactText, []byte(" \n")...)
externalMetricFamilyWithInvalidLabelValue := &dto.MetricFamily{
Name: proto.String("name"),
Help: proto.String("docstring"),
Type: dto.MetricType_COUNTER.Enum(),
Metric: []*dto.Metric{
{
Label: []*dto.LabelPair{
{
Name: proto.String("constname"),
Value: proto.String("\xFF"),
},
{
Name: proto.String("labelname"),
Value: proto.String("different_val"),
},
},
Counter: &dto.Counter{
Value: proto.Float64(42),
},
},
},
}
expectedMetricFamilyInvalidLabelValueAsText := []byte(`An error has occurred while serving metrics:
collected metric "name" { label:<name:"constname" value:"\377" > label:<name:"labelname" value:"different_val" > counter:<value:42 > } has a label named "constname" whose value is not utf8: "\xff"
`)
summary := prometheus.NewSummary(prometheus.SummaryOpts{
Name: "complex",
Help: "A metric to check collisions with _sum and _count.",
Objectives: map[float64]float64{0.5: 0.05, 0.9: 0.01, 0.99: 0.001},
})
summaryAsText := []byte(`# HELP complex A metric to check collisions with _sum and _count.
# TYPE complex summary
complex{quantile="0.5"} NaN
complex{quantile="0.9"} NaN
complex{quantile="0.99"} NaN
complex_sum 0
complex_count 0
`)
histogram := prometheus.NewHistogram(prometheus.HistogramOpts{
Name: "complex",
Help: "A metric to check collisions with _sun, _count, and _bucket.",
})
externalMetricFamilyWithBucketSuffix := &dto.MetricFamily{
Name: proto.String("complex_bucket"),
Help: proto.String("externaldocstring"),
Type: dto.MetricType_COUNTER.Enum(),
Metric: []*dto.Metric{
{
Counter: &dto.Counter{
Value: proto.Float64(1),
},
},
},
}
externalMetricFamilyWithBucketSuffixAsText := []byte(`# HELP complex_bucket externaldocstring
# TYPE complex_bucket counter
complex_bucket 1
`)
externalMetricFamilyWithCountSuffix := &dto.MetricFamily{
Name: proto.String("complex_count"),
Help: proto.String("externaldocstring"),
Type: dto.MetricType_COUNTER.Enum(),
Metric: []*dto.Metric{
{
Counter: &dto.Counter{
Value: proto.Float64(1),
},
},
},
}
bucketCollisionMsg := []byte(`An error has occurred while serving metrics:
collected metric named "complex_bucket" collides with previously collected histogram named "complex"
`)
summaryCountCollisionMsg := []byte(`An error has occurred while serving metrics:
collected metric named "complex_count" collides with previously collected summary named "complex"
`)
histogramCountCollisionMsg := []byte(`An error has occurred while serving metrics:
collected metric named "complex_count" collides with previously collected histogram named "complex"
`)
externalMetricFamilyWithDuplicateLabel := &dto.MetricFamily{
Name: proto.String("broken_metric"),
Help: proto.String("The registry should detect the duplicate label."),
Type: dto.MetricType_COUNTER.Enum(),
Metric: []*dto.Metric{
{
Label: []*dto.LabelPair{
{
Name: proto.String("foo"),
Value: proto.String("bar"),
},
{
Name: proto.String("foo"),
Value: proto.String("baz"),
},
},
Counter: &dto.Counter{
Value: proto.Float64(2.7),
},
},
},
}
duplicateLabelMsg := []byte(`An error has occurred while serving metrics:
collected metric "broken_metric" { label:<name:"foo" value:"bar" > label:<name:"foo" value:"baz" > counter:<value:2.7 > } has two or more labels with the same name: foo
`)
type output struct {
headers map[string]string
body []byte
}
scenarios := []struct {
headers map[string]string
out output
collector prometheus.Collector
externalMF []*dto.MetricFamily
}{
{ // 0
headers: map[string]string{
"Accept": "foo/bar;q=0.2, dings/bums;q=0.8",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; version=0.0.4; charset=utf-8`,
},
body: []byte{},
},
},
{ // 1
headers: map[string]string{
"Accept": "foo/bar;q=0.2, application/quark;q=0.8",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; version=0.0.4; charset=utf-8`,
},
body: []byte{},
},
},
{ // 2
headers: map[string]string{
"Accept": "foo/bar;q=0.2, application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=bla;q=0.8",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; version=0.0.4; charset=utf-8`,
},
body: []byte{},
},
},
{ // 3
headers: map[string]string{
"Accept": "text/plain;q=0.2, application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=delimited;q=0.8",
},
out: output{
headers: map[string]string{
"Content-Type": `application/vnd.google.protobuf; proto=io.prometheus.client.MetricFamily; encoding=delimited`,
},
body: []byte{},
},
},
{ // 4
headers: map[string]string{
"Accept": "application/json",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; version=0.0.4; charset=utf-8`,
},
body: expectedMetricFamilyAsText,
},
collector: metricVec,
},
{ // 5
headers: map[string]string{
"Accept": "application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=delimited",
},
out: output{
headers: map[string]string{
"Content-Type": `application/vnd.google.protobuf; proto=io.prometheus.client.MetricFamily; encoding=delimited`,
},
body: expectedMetricFamilyAsBytes,
},
collector: metricVec,
},
{ // 6
headers: map[string]string{
"Accept": "application/json",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; version=0.0.4; charset=utf-8`,
},
body: externalMetricFamilyAsText,
},
externalMF: []*dto.MetricFamily{externalMetricFamily},
},
{ // 7
headers: map[string]string{
"Accept": "application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=delimited",
},
out: output{
headers: map[string]string{
"Content-Type": `application/vnd.google.protobuf; proto=io.prometheus.client.MetricFamily; encoding=delimited`,
},
body: externalMetricFamilyAsBytes,
},
externalMF: []*dto.MetricFamily{externalMetricFamily},
},
{ // 8
headers: map[string]string{
"Accept": "application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=delimited",
},
out: output{
headers: map[string]string{
"Content-Type": `application/vnd.google.protobuf; proto=io.prometheus.client.MetricFamily; encoding=delimited`,
},
body: bytes.Join(
[][]byte{
externalMetricFamilyAsBytes,
expectedMetricFamilyAsBytes,
},
[]byte{},
),
},
collector: metricVec,
externalMF: []*dto.MetricFamily{externalMetricFamily},
},
{ // 9
headers: map[string]string{
"Accept": "text/plain",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; version=0.0.4; charset=utf-8`,
},
body: []byte{},
},
},
{ // 10
headers: map[string]string{
"Accept": "application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=bla;q=0.2, text/plain;q=0.5",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; version=0.0.4; charset=utf-8`,
},
body: expectedMetricFamilyAsText,
},
collector: metricVec,
},
{ // 11
headers: map[string]string{
"Accept": "application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=bla;q=0.2, text/plain;q=0.5;version=0.0.4",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; version=0.0.4; charset=utf-8`,
},
body: bytes.Join(
[][]byte{
externalMetricFamilyAsText,
expectedMetricFamilyAsText,
},
[]byte{},
),
},
collector: metricVec,
externalMF: []*dto.MetricFamily{externalMetricFamily},
},
{ // 12
headers: map[string]string{
"Accept": "application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=delimited;q=0.2, text/plain;q=0.5;version=0.0.2",
},
out: output{
headers: map[string]string{
"Content-Type": `application/vnd.google.protobuf; proto=io.prometheus.client.MetricFamily; encoding=delimited`,
},
body: bytes.Join(
[][]byte{
externalMetricFamilyAsBytes,
expectedMetricFamilyAsBytes,
},
[]byte{},
),
},
collector: metricVec,
externalMF: []*dto.MetricFamily{externalMetricFamily},
},
{ // 13
headers: map[string]string{
"Accept": "application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=text;q=0.5, application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=delimited;q=0.4",
},
out: output{
headers: map[string]string{
"Content-Type": `application/vnd.google.protobuf; proto=io.prometheus.client.MetricFamily; encoding=text`,
},
body: bytes.Join(
[][]byte{
externalMetricFamilyAsProtoText,
expectedMetricFamilyAsProtoText,
},
[]byte{},
),
},
collector: metricVec,
externalMF: []*dto.MetricFamily{externalMetricFamily},
},
{ // 14
headers: map[string]string{
"Accept": "application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=compact-text",
},
out: output{
headers: map[string]string{
"Content-Type": `application/vnd.google.protobuf; proto=io.prometheus.client.MetricFamily; encoding=compact-text`,
},
body: bytes.Join(
[][]byte{
externalMetricFamilyAsProtoCompactText,
expectedMetricFamilyAsProtoCompactText,
},
[]byte{},
),
},
collector: metricVec,
externalMF: []*dto.MetricFamily{externalMetricFamily},
},
{ // 15
headers: map[string]string{
"Accept": "application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=compact-text",
},
out: output{
headers: map[string]string{
"Content-Type": `application/vnd.google.protobuf; proto=io.prometheus.client.MetricFamily; encoding=compact-text`,
},
body: bytes.Join(
[][]byte{
externalMetricFamilyAsProtoCompactText,
expectedMetricFamilyMergedWithExternalAsProtoCompactText,
},
[]byte{},
),
},
collector: metricVec,
externalMF: []*dto.MetricFamily{
externalMetricFamily,
externalMetricFamilyWithSameName,
},
},
{ // 16
headers: map[string]string{
"Accept": "application/vnd.google.protobuf;proto=io.prometheus.client.MetricFamily;encoding=compact-text",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; charset=utf-8`,
},
body: expectedMetricFamilyInvalidLabelValueAsText,
},
collector: metricVec,
externalMF: []*dto.MetricFamily{
externalMetricFamily,
externalMetricFamilyWithInvalidLabelValue,
},
},
{ // 17
headers: map[string]string{
"Accept": "text/plain",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; version=0.0.4; charset=utf-8`,
},
body: expectedMetricFamilyAsText,
},
collector: uncheckedCollector{metricVec},
},
{ // 18
headers: map[string]string{
"Accept": "text/plain",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; charset=utf-8`,
},
body: histogramCountCollisionMsg,
},
collector: histogram,
externalMF: []*dto.MetricFamily{
externalMetricFamilyWithCountSuffix,
},
},
{ // 19
headers: map[string]string{
"Accept": "text/plain",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; charset=utf-8`,
},
body: bucketCollisionMsg,
},
collector: histogram,
externalMF: []*dto.MetricFamily{
externalMetricFamilyWithBucketSuffix,
},
},
{ // 20
headers: map[string]string{
"Accept": "text/plain",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; charset=utf-8`,
},
body: summaryCountCollisionMsg,
},
collector: summary,
externalMF: []*dto.MetricFamily{
externalMetricFamilyWithCountSuffix,
},
},
{ // 21
headers: map[string]string{
"Accept": "text/plain",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; version=0.0.4; charset=utf-8`,
},
body: bytes.Join(
[][]byte{
summaryAsText,
externalMetricFamilyWithBucketSuffixAsText,
},
[]byte{},
),
},
collector: summary,
externalMF: []*dto.MetricFamily{
externalMetricFamilyWithBucketSuffix,
},
},
{ // 22
headers: map[string]string{
"Accept": "text/plain",
},
out: output{
headers: map[string]string{
"Content-Type": `text/plain; charset=utf-8`,
},
body: duplicateLabelMsg,
},
externalMF: []*dto.MetricFamily{
externalMetricFamilyWithDuplicateLabel,
},
},
}
for i, scenario := range scenarios {
registry := prometheus.NewPedanticRegistry()
gatherer := prometheus.Gatherer(registry)
if scenario.externalMF != nil {
gatherer = prometheus.Gatherers{
registry,
prometheus.GathererFunc(func() ([]*dto.MetricFamily, error) {
return scenario.externalMF, 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
if scenario.collector != nil {
registry.MustRegister(scenario.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
writer := httptest.NewRecorder()
handler := promhttp.HandlerFor(gatherer, promhttp.HandlerOpts{})
request, _ := http.NewRequest("GET", "/", nil)
for key, value := range scenario.headers {
request.Header.Add(key, value)
}
handler.ServeHTTP(writer, request)
for key, value := range scenario.out.headers {
if writer.Header().Get(key) != value {
t.Errorf(
"%d. expected %q for header %q, got %q",
i, value, key, writer.Header().Get(key),
)
}
}
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
if !bytes.Equal(scenario.out.body, writer.Body.Bytes()) {
t.Errorf(
"%d. expected body:\n%s\ngot body:\n%s\n",
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
i, scenario.out.body, writer.Body.Bytes(),
)
}
}
}
func TestHandler(t *testing.T) {
testHandler(t)
}
func BenchmarkHandler(b *testing.B) {
for i := 0; i < b.N; i++ {
testHandler(b)
}
}
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 TestAlreadyRegistered(t *testing.T) {
original := prometheus.NewCounterVec(
prometheus.CounterOpts{
Name: "test",
Help: "help",
ConstLabels: prometheus.Labels{"const": "label"},
},
[]string{"foo", "bar"},
)
equalButNotSame := prometheus.NewCounterVec(
prometheus.CounterOpts{
Name: "test",
Help: "help",
ConstLabels: prometheus.Labels{"const": "label"},
},
[]string{"foo", "bar"},
)
originalWithoutConstLabel := prometheus.NewCounterVec(
prometheus.CounterOpts{
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
Name: "test",
Help: "help",
},
[]string{"foo", "bar"},
)
equalButNotSameWithoutConstLabel := prometheus.NewCounterVec(
prometheus.CounterOpts{
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
Name: "test",
Help: "help",
},
[]string{"foo", "bar"},
)
scenarios := []struct {
name string
originalCollector prometheus.Collector
registerWith func(prometheus.Registerer) prometheus.Registerer
newCollector prometheus.Collector
reRegisterWith func(prometheus.Registerer) prometheus.Registerer
}{
{
"RegisterNormallyReregisterNormally",
original,
func(r prometheus.Registerer) prometheus.Registerer { return r },
equalButNotSame,
func(r prometheus.Registerer) prometheus.Registerer { return r },
},
{
"RegisterNormallyReregisterWrapped",
original,
func(r prometheus.Registerer) prometheus.Registerer { return r },
equalButNotSameWithoutConstLabel,
func(r prometheus.Registerer) prometheus.Registerer {
return prometheus.WrapRegistererWith(prometheus.Labels{"const": "label"}, r)
},
},
{
"RegisterWrappedReregisterWrapped",
originalWithoutConstLabel,
func(r prometheus.Registerer) prometheus.Registerer {
return prometheus.WrapRegistererWith(prometheus.Labels{"const": "label"}, r)
},
equalButNotSameWithoutConstLabel,
func(r prometheus.Registerer) prometheus.Registerer {
return prometheus.WrapRegistererWith(prometheus.Labels{"const": "label"}, r)
},
},
{
"RegisterWrappedReregisterNormally",
originalWithoutConstLabel,
func(r prometheus.Registerer) prometheus.Registerer {
return prometheus.WrapRegistererWith(prometheus.Labels{"const": "label"}, r)
},
equalButNotSame,
func(r prometheus.Registerer) prometheus.Registerer { return r },
},
{
"RegisterDoublyWrappedReregisterDoublyWrapped",
originalWithoutConstLabel,
func(r prometheus.Registerer) prometheus.Registerer {
return prometheus.WrapRegistererWithPrefix(
"wrap_",
prometheus.WrapRegistererWith(prometheus.Labels{"const": "label"}, r),
)
},
equalButNotSameWithoutConstLabel,
func(r prometheus.Registerer) prometheus.Registerer {
return prometheus.WrapRegistererWithPrefix(
"wrap_",
prometheus.WrapRegistererWith(prometheus.Labels{"const": "label"}, r),
)
},
},
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 _, s := range scenarios {
t.Run(s.name, func(t *testing.T) {
var err error
reg := prometheus.NewRegistry()
if err = s.registerWith(reg).Register(s.originalCollector); err != nil {
t.Fatal(err)
}
if err = s.reRegisterWith(reg).Register(s.newCollector); err == nil {
t.Fatal("expected error when registering new collector")
}
are := &prometheus.AlreadyRegisteredError{}
if errors.As(err, are) {
if are.ExistingCollector != s.originalCollector {
t.Error("expected original collector but got something else")
}
if are.ExistingCollector == s.newCollector {
t.Error("expected original collector but got new one")
}
} else {
t.Error("unexpected error:", 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
}
}
// TestRegisterUnregisterCollector ensures registering and unregistering a
// collector doesn't leave any dangling metrics.
// We use NewGoCollector as a nice concrete example of a collector with
// multiple metrics.
func TestRegisterUnregisterCollector(t *testing.T) {
col := prometheus.NewGoCollector()
reg := prometheus.NewRegistry()
reg.MustRegister(col)
reg.Unregister(col)
if metrics, err := reg.Gather(); err != nil {
t.Error("error gathering sample metric")
} else if len(metrics) != 0 {
t.Error("should have unregistered metric")
}
}
// TestHistogramVecRegisterGatherConcurrency is an end-to-end test that
// concurrently calls Observe on random elements of a HistogramVec while the
// same HistogramVec is registered concurrently and the Gather method of the
// registry is called concurrently.
func TestHistogramVecRegisterGatherConcurrency(t *testing.T) {
labelNames := make([]string, 16) // Need at least 13 to expose #512.
for i := range labelNames {
labelNames[i] = fmt.Sprint("label_", i)
}
var (
reg = prometheus.NewPedanticRegistry()
hv = prometheus.NewHistogramVec(
prometheus.HistogramOpts{
Name: "test_histogram",
Help: "This helps testing.",
ConstLabels: prometheus.Labels{"foo": "bar"},
},
labelNames,
)
labelValues = []string{"a", "b", "c", "alpha", "beta", "gamma", "aleph", "beth", "gimel"}
quit = make(chan struct{})
wg sync.WaitGroup
)
observe := func() {
defer wg.Done()
for {
select {
case <-quit:
return
default:
obs := rand.NormFloat64()*.1 + .2
values := make([]string, 0, len(labelNames))
for range labelNames {
values = append(values, labelValues[rand.Intn(len(labelValues))])
}
hv.WithLabelValues(values...).Observe(obs)
}
}
}
register := func() {
defer wg.Done()
for {
select {
case <-quit:
return
default:
if err := reg.Register(hv); err != nil {
if !errors.As(err, &prometheus.AlreadyRegisteredError{}) {
t.Error("Registering failed:", err)
}
}
time.Sleep(7 * time.Millisecond)
}
}
}
gather := func() {
defer wg.Done()
for {
select {
case <-quit:
return
default:
if g, err := reg.Gather(); err != nil {
t.Error("Gathering failed:", err)
} else {
if len(g) == 0 {
continue
}
if len(g) != 1 {
t.Error("Gathered unexpected number of metric families:", len(g))
}
if len(g[0].Metric[0].Label) != len(labelNames)+1 {
t.Error("Gathered unexpected number of label pairs:", len(g[0].Metric[0].Label))
}
}
time.Sleep(4 * time.Millisecond)
}
}
}
wg.Add(10)
go observe()
go observe()
go register()
go observe()
go gather()
go observe()
go register()
go observe()
go gather()
go observe()
time.Sleep(time.Second)
close(quit)
wg.Wait()
}
func TestWriteToTextfile(t *testing.T) {
expectedOut := `# HELP test_counter test counter
# TYPE test_counter counter
test_counter{name="qux"} 1
# HELP test_gauge test gauge
# TYPE test_gauge gauge
test_gauge{name="baz"} 1.1
# HELP test_hist test histogram
# TYPE test_hist histogram
test_hist_bucket{name="bar",le="0.005"} 0
test_hist_bucket{name="bar",le="0.01"} 0
test_hist_bucket{name="bar",le="0.025"} 0
test_hist_bucket{name="bar",le="0.05"} 0
test_hist_bucket{name="bar",le="0.1"} 0
test_hist_bucket{name="bar",le="0.25"} 0
test_hist_bucket{name="bar",le="0.5"} 0
test_hist_bucket{name="bar",le="1"} 1
test_hist_bucket{name="bar",le="2.5"} 1
test_hist_bucket{name="bar",le="5"} 2
test_hist_bucket{name="bar",le="10"} 2
test_hist_bucket{name="bar",le="+Inf"} 2
test_hist_sum{name="bar"} 3.64
test_hist_count{name="bar"} 2
# HELP test_summary test summary
# TYPE test_summary summary
test_summary{name="foo",quantile="0.5"} 10
test_summary{name="foo",quantile="0.9"} 20
test_summary{name="foo",quantile="0.99"} 20
test_summary_sum{name="foo"} 30
test_summary_count{name="foo"} 2
`
registry := prometheus.NewRegistry()
summary := prometheus.NewSummaryVec(
prometheus.SummaryOpts{
Name: "test_summary",
Help: "test summary",
Objectives: map[float64]float64{
0.5: 0.05,
0.9: 0.01,
0.99: 0.001,
},
},
[]string{"name"},
)
histogram := prometheus.NewHistogramVec(
prometheus.HistogramOpts{
Name: "test_hist",
Help: "test histogram",
},
[]string{"name"},
)
gauge := prometheus.NewGaugeVec(
prometheus.GaugeOpts{
Name: "test_gauge",
Help: "test gauge",
},
[]string{"name"},
)
counter := prometheus.NewCounterVec(
prometheus.CounterOpts{
Name: "test_counter",
Help: "test counter",
},
[]string{"name"},
)
registry.MustRegister(summary)
registry.MustRegister(histogram)
registry.MustRegister(gauge)
registry.MustRegister(counter)
summary.With(prometheus.Labels{"name": "foo"}).Observe(10)
summary.With(prometheus.Labels{"name": "foo"}).Observe(20)
histogram.With(prometheus.Labels{"name": "bar"}).Observe(0.93)
histogram.With(prometheus.Labels{"name": "bar"}).Observe(2.71)
gauge.With(prometheus.Labels{"name": "baz"}).Set(1.1)
counter.With(prometheus.Labels{"name": "qux"}).Inc()
tmpfile, err := os.CreateTemp("", "prom_registry_test")
if err != nil {
t.Fatal(err)
}
defer os.Remove(tmpfile.Name())
if err := prometheus.WriteToTextfile(tmpfile.Name(), registry); err != nil {
t.Fatal(err)
}
fileBytes, err := os.ReadFile(tmpfile.Name())
if err != nil {
t.Fatal(err)
}
fileContents := string(fileBytes)
if fileContents != expectedOut {
t.Errorf(
"files don't match, got:\n%s\nwant:\n%s",
fileContents, expectedOut,
)
}
}
// collidingCollector is a collection of prometheus.Collectors,
// and is itself a prometheus.Collector.
type collidingCollector struct {
i int
name string
a, b, c, d prometheus.Collector
}
// Describe satisifies part of the prometheus.Collector interface.
func (m *collidingCollector) Describe(desc chan<- *prometheus.Desc) {
m.a.Describe(desc)
m.b.Describe(desc)
m.c.Describe(desc)
m.d.Describe(desc)
}
// Collect satisifies part of the prometheus.Collector interface.
func (m *collidingCollector) Collect(metric chan<- prometheus.Metric) {
m.a.Collect(metric)
m.b.Collect(metric)
m.c.Collect(metric)
m.d.Collect(metric)
}
// TestAlreadyRegistered will fail with the old, weaker hash function. It is
// taken from https://play.golang.org/p/HpV7YE6LI_4 , authored by @awilliams.
func TestAlreadyRegisteredCollision(t *testing.T) {
reg := prometheus.NewRegistry()
for i := 0; i < 10000; i++ {
// A collector should be considered unique if its name and const
// label values are unique.
name := fmt.Sprintf("test-collector-%010d", i)
collector := collidingCollector{
i: i,
name: name,
a: prometheus.NewCounter(prometheus.CounterOpts{
Name: "my_collector_a",
ConstLabels: prometheus.Labels{
"name": name,
"type": "test",
},
}),
b: prometheus.NewCounter(prometheus.CounterOpts{
Name: "my_collector_b",
ConstLabels: prometheus.Labels{
"name": name,
"type": "test",
},
}),
c: prometheus.NewCounter(prometheus.CounterOpts{
Name: "my_collector_c",
ConstLabels: prometheus.Labels{
"name": name,
"type": "test",
},
}),
d: prometheus.NewCounter(prometheus.CounterOpts{
Name: "my_collector_d",
ConstLabels: prometheus.Labels{
"name": name,
"type": "test",
},
}),
}
// Register should not fail, since each collector has a unique
// set of sub-collectors, determined by their names and const label values.
if err := reg.Register(&collector); err != nil {
are := &prometheus.AlreadyRegisteredError{}
if !errors.As(err, are) {
t.Fatal(err)
}
previous := are.ExistingCollector.(*collidingCollector)
current := are.NewCollector.(*collidingCollector)
t.Errorf("Unexpected registration error: %q\nprevious collector: %s (i=%d)\ncurrent collector %s (i=%d)", are, previous.name, previous.i, current.name, current.i)
}
}
}
Added Transactional Gatherer allowed cached solutions (#989) * Added cached collector. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> update. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Attempt 2 Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Added blocking registry, with raw collector and transactional handler. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Added fast path to normal (empty) registry to save 8 allocs and 3K5B per Gather. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Simplified API, added tests. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Fix. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Simplified implementation. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Added benchmark. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Optimized. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Optimization attempt. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Revert "Optimization attempt." This reverts commit 2fcaf51be9a12b4b95413b6b3e0c13fabfaaf73f. Optimization was not worth it: benchstat v1.txt v2.txt name old time/op new time/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 2.64µs ± 0% 4.05µs ± 0% ~ (p=1.000 n=1+1) CachedTGatherer_Update/Update_of_all_elements_with_reset-12 701ms ± 0% 358ms ± 0% ~ (p=1.000 n=1+1) CachedTGatherer_Update/Gather-12 535µs ± 0% 703934µs ± 0% ~ (p=1.000 n=1+1) name old alloc/op new alloc/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 208B ± 0% 208B ± 0% ~ (all equal) CachedTGatherer_Update/Update_of_all_elements_with_reset-12 40.2MB ± 0% 41.1MB ± 0% ~ (p=1.000 n=1+1) CachedTGatherer_Update/Gather-12 48.6kB ± 0% 84.3kB ± 0% ~ (p=1.000 n=1+1) name old allocs/op new allocs/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 3.00 ± 0% 3.00 ± 0% ~ (all equal) CachedTGatherer_Update/Update_of_all_elements_with_reset-12 6.00 ± 0% 4003.00 ± 0% ~ (p=1.000 n=1+1) CachedTGatherer_Update/Gather-12 1.00k ± 0% 2.01k ± 0% ~ (p=1.000 n=1+1) * nit. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Another optimization attempt. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * rename and further optimization. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Hopefully final optimization. benchstat -delta-test=none v6.txt v9.txt name old time/op new time/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 13.1ms ± 0% 0.0ms ± 0% -99.81% CachedTGatherer_Update/Update_of_all_elements_with_reset-12 309ms ± 0% 282ms ± 0% -8.77% CachedTGatherer_Update/Gather-12 422ms ± 0% 0ms ± 0% -99.95% name old alloc/op new alloc/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 208B ± 0% 208B ± 0% 0.00% CachedTGatherer_Update/Update_of_all_elements_with_reset-12 2.47kB ± 0% 1.67kB ± 0% -32.56% CachedTGatherer_Update/Gather-12 52.8kB ± 0% 24.6kB ± 0% -53.34% name old allocs/op new allocs/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 3.00 ± 0% 3.00 ± 0% 0.00% CachedTGatherer_Update/Update_of_all_elements_with_reset-12 0.00 0.00 0.00% CachedTGatherer_Update/Gather-12 1.00k ± 0% 0.00k ± 0% -99.60% Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Removed obsolete comment Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Fixed tests. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Removed cache. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Fixed tests. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Re-add cache. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Removed cache. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com>
2022-02-23 14:22:52 +03:00
type tGatherer struct {
done bool
err error
}
func (g *tGatherer) Gather() (_ []*dto.MetricFamily, done func(), err error) {
name := "g1"
val := 1.0
return []*dto.MetricFamily{
{Name: &name, Metric: []*dto.Metric{{Gauge: &dto.Gauge{Value: &val}}}},
}, func() { g.done = true }, g.err
}
func TestNewMultiTRegistry(t *testing.T) {
treg := &tGatherer{}
t.Run("one registry", func(t *testing.T) {
m := prometheus.NewMultiTRegistry(treg)
ret, done, err := m.Gather()
if err != nil {
t.Error("gather failed:", err)
}
done()
if len(ret) != 1 {
t.Error("unexpected number of metric families, expected 1, got", ret)
}
if !treg.done {
t.Error("inner transactional registry not marked as done")
}
})
reg := prometheus.NewRegistry()
if err := reg.Register(prometheus.NewCounter(prometheus.CounterOpts{Name: "c1", Help: "help c1"})); err != nil {
t.Error("registration failed:", err)
}
// Note on purpose two registries will have exactly same metric family name (but with different string).
// This behaviour is undefined at the moment.
if err := reg.Register(prometheus.NewGauge(prometheus.GaugeOpts{Name: "g1", Help: "help g1"})); err != nil {
t.Error("registration failed:", err)
}
treg.done = false
t.Run("two registries", func(t *testing.T) {
m := prometheus.NewMultiTRegistry(prometheus.ToTransactionalGatherer(reg), treg)
ret, done, err := m.Gather()
if err != nil {
t.Error("gather failed:", err)
}
done()
if len(ret) != 3 {
t.Error("unexpected number of metric families, expected 3, got", ret)
}
if !treg.done {
t.Error("inner transactional registry not marked as done")
}
})
treg.done = false
// Inject error.
treg.err = errors.New("test err")
t.Run("two registries, one with error", func(t *testing.T) {
m := prometheus.NewMultiTRegistry(prometheus.ToTransactionalGatherer(reg), treg)
ret, done, err := m.Gather()
if !errors.Is(err, treg.err) {
Added Transactional Gatherer allowed cached solutions (#989) * Added cached collector. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> update. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Attempt 2 Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Added blocking registry, with raw collector and transactional handler. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Added fast path to normal (empty) registry to save 8 allocs and 3K5B per Gather. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Simplified API, added tests. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Fix. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Simplified implementation. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Added benchmark. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> Optimized. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Optimization attempt. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Revert "Optimization attempt." This reverts commit 2fcaf51be9a12b4b95413b6b3e0c13fabfaaf73f. Optimization was not worth it: benchstat v1.txt v2.txt name old time/op new time/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 2.64µs ± 0% 4.05µs ± 0% ~ (p=1.000 n=1+1) CachedTGatherer_Update/Update_of_all_elements_with_reset-12 701ms ± 0% 358ms ± 0% ~ (p=1.000 n=1+1) CachedTGatherer_Update/Gather-12 535µs ± 0% 703934µs ± 0% ~ (p=1.000 n=1+1) name old alloc/op new alloc/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 208B ± 0% 208B ± 0% ~ (all equal) CachedTGatherer_Update/Update_of_all_elements_with_reset-12 40.2MB ± 0% 41.1MB ± 0% ~ (p=1.000 n=1+1) CachedTGatherer_Update/Gather-12 48.6kB ± 0% 84.3kB ± 0% ~ (p=1.000 n=1+1) name old allocs/op new allocs/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 3.00 ± 0% 3.00 ± 0% ~ (all equal) CachedTGatherer_Update/Update_of_all_elements_with_reset-12 6.00 ± 0% 4003.00 ± 0% ~ (p=1.000 n=1+1) CachedTGatherer_Update/Gather-12 1.00k ± 0% 2.01k ± 0% ~ (p=1.000 n=1+1) * nit. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Another optimization attempt. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * rename and further optimization. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Hopefully final optimization. benchstat -delta-test=none v6.txt v9.txt name old time/op new time/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 13.1ms ± 0% 0.0ms ± 0% -99.81% CachedTGatherer_Update/Update_of_all_elements_with_reset-12 309ms ± 0% 282ms ± 0% -8.77% CachedTGatherer_Update/Gather-12 422ms ± 0% 0ms ± 0% -99.95% name old alloc/op new alloc/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 208B ± 0% 208B ± 0% 0.00% CachedTGatherer_Update/Update_of_all_elements_with_reset-12 2.47kB ± 0% 1.67kB ± 0% -32.56% CachedTGatherer_Update/Gather-12 52.8kB ± 0% 24.6kB ± 0% -53.34% name old allocs/op new allocs/op delta CachedTGatherer_Update/Update_of_one_element_without_reset-12 3.00 ± 0% 3.00 ± 0% 0.00% CachedTGatherer_Update/Update_of_all_elements_with_reset-12 0.00 0.00 0.00% CachedTGatherer_Update/Gather-12 1.00k ± 0% 0.00k ± 0% -99.60% Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Removed obsolete comment Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Fixed tests. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Removed cache. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Fixed tests. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Re-add cache. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com> * Removed cache. Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com>
2022-02-23 14:22:52 +03:00
t.Error("unexpected error:", err)
}
done()
if len(ret) != 3 {
t.Error("unexpected number of metric families, expected 3, got", ret)
}
// Still on error, we expect done to be triggered.
if !treg.done {
t.Error("inner transactional registry not marked as done")
}
})
}
// This example shows how to use multiple registries for registering and
// unregistering groups of metrics.
func ExampleRegistry_grouping() {
// Create a global registry.
globalReg := prometheus.NewRegistry()
// Spawn 10 workers, each of which will have their own group of metrics.
for i := 0; i < 10; i++ {
// Create a new registry for each worker, which acts as a group of
// worker-specific metrics.
workerReg := prometheus.NewRegistry()
globalReg.Register(workerReg)
go func(workerID int) {
// Once the worker is done, it can unregister itself.
defer globalReg.Unregister(workerReg)
workTime := prometheus.NewCounter(prometheus.CounterOpts{
Name: "worker_total_work_time_milliseconds",
ConstLabels: prometheus.Labels{
// Generate a label unique to this worker so its metric doesn't
// collide with the metrics from other workers.
"worker_id": fmt.Sprintf("%d", workerID),
},
})
workerReg.MustRegister(workTime)
start := time.Now()
time.Sleep(time.Millisecond * time.Duration(rand.Intn(100)))
workTime.Add(float64(time.Since(start).Milliseconds()))
}(i)
}
}
type customCollector struct {
collectFunc func(ch chan<- prometheus.Metric)
}
func (co *customCollector) Describe(_ chan<- *prometheus.Desc) {}
func (co *customCollector) Collect(ch chan<- prometheus.Metric) {
co.collectFunc(ch)
}
// TestCheckMetricConsistency
func TestCheckMetricConsistency(t *testing.T) {
reg := prometheus.NewRegistry()
timestamp := time.Now()
desc := prometheus.NewDesc("metric_a", "", nil, nil)
metric := prometheus.MustNewConstMetric(desc, prometheus.CounterValue, 1)
validCollector := &customCollector{
collectFunc: func(ch chan<- prometheus.Metric) {
ch <- prometheus.NewMetricWithTimestamp(timestamp.Add(-1*time.Minute), metric)
ch <- prometheus.NewMetricWithTimestamp(timestamp, metric)
},
}
reg.MustRegister(validCollector)
_, err := reg.Gather()
if err != nil {
t.Error("metric validation should succeed:", err)
}
reg.Unregister(validCollector)
invalidCollector := &customCollector{
collectFunc: func(ch chan<- prometheus.Metric) {
ch <- prometheus.NewMetricWithTimestamp(timestamp, metric)
ch <- prometheus.NewMetricWithTimestamp(timestamp, metric)
},
}
reg.MustRegister(invalidCollector)
_, err = reg.Gather()
if err == nil {
t.Error("metric validation should return an error")
}
reg.Unregister(invalidCollector)
}