Commit Graph

263 Commits

Author SHA1 Message Date
tidwall 7ff0d18868 Change hooks collection type from hashmap to btree
This commit changes the collection type that holds all of the
hooks from a hashmap to a btree. This allows for better
flexibility for operations that need to perform range searches
and scanning of the collection.
2021-09-29 07:19:00 -07:00
tidwall decafae2d7 Update hook expiration logic 2021-09-29 07:19:00 -07:00
tidwall 7e10a80319 Return hook ttl with HOOKS request 2021-09-29 07:19:00 -07:00
tidwall d38aed4d2b Increase the precision of TIMEOUT
This commit ensures that the TIMEOUT is always checked prior to
returning data to the client, and that the elapsed command time
cannot be greater than the timeout value.
2021-09-29 07:19:00 -07:00
tidwall 9b760ffdd5 Add unix socket support 2021-09-29 07:19:00 -07:00
tidwall f9e6f74753 Use struct for server options 2021-09-29 07:19:00 -07:00
Benjamin Ramser 6eb1cca78c fix: point -> circle fallthrough 2021-09-05 12:05:33 +02:00
Benjamin Ramser bc62edb692 feat: add sector 2021-09-05 11:48:34 +02:00
tidwall c8389fe52c Fix memory leak with group id
This commit fixes a memory leak that was being caused by hooks
hanging on to the geofence group ids past the life of the object.
2021-08-20 05:00:14 -07:00
tidwall 694ad1fb62 Update btree/buntdb 2021-07-31 07:42:58 -07:00
tidwall 401670e621 Fix NEARBY with SPARSE returning too many results
fixes #618
2021-07-22 08:39:57 -07:00
tidwall 9e68703841 Update expiration logic
This commit changes the logic for managing the expiration of
objects in the database.

Before: There was a server-wide hashmap that stored the
collection key, id, and expiration timestamp for all objects
that had a TTL. The hashmap was occasionally probed at 20
random positions, looking for objects that have expired. Those
expired objects were immediately deleted, and if there was 5
or more objects deleted, then the probe happened again, with
no delay. If the number of objects was less than 5 then the
there was a 1/10th of a second delay before the next probe.

Now: Rather than a server-wide hashmap, each collection has
its own ordered priority queue that stores objects with TTLs.
Rather than probing, there is a background routine that
executes every 1/10th of a second, which pops the expired
objects from the collection queues, and deletes them.

The collection/queue method is a more stable approach than
the hashmap/probing method. With probing, we can run into
major cache misses for some cases where there is wide
TTL duration, such as in the hours or days. This may cause
the system to occasionally fall behind, leaving should-be
expired objects in memory. Using a queue, there is no
cache misses, all objects that should be expired will be
right away, regardless of the TTL durations.

Fixes #616
2021-07-12 13:37:50 -07:00
tidwall dd4d31ae1b Fix last merge 2021-07-11 10:09:51 -07:00
tidwall 579a41abae Merge branch 'housecanary-fix-knn' 2021-07-11 10:02:59 -07:00
tidwall 0d83b1ca53 Merge branch 'rshura-optimize-field-match' 2021-07-10 17:59:22 -07:00
Josh 1467cba769
Merge pull request #552 from rshura/clip-by
Add CLIPBY subcommand to INTERSECTS/WITHIN
2021-07-10 09:24:57 -07:00
tidwall d2f747933f clear linter messages 2021-07-10 03:59:27 -07:00
tidwall df8d3d7b12 Close follower files before finishing aofshrink
fixes #449
2021-06-13 07:53:27 -07:00
tidwall af43d5a7c7 Fix invalid healthz output 2021-06-09 10:55:07 -07:00
tidwall 27335c37f1 Always OK for leader 2021-05-25 16:42:26 -07:00
tidwall 30c3e81e0d Added HEALTHZ command
Returns 'ok' if the server is the leader or a follower with
a 'caught up' log.

This is mainly for HTTP connections that are using an
orchestration environment like kubernetes, but will work as a
general RESP command.

For HTTP a '200 OK' for 'caught up' and
'500 Internal Server Error' otherwise.

See #608
2021-05-25 16:36:49 -07:00
Oliver 6b0282d960 add more metrics, add metrics index page 2021-05-13 21:14:05 -04:00
Oliver 7ec4d58392 cleanup / fix tests 2021-05-13 21:14:05 -04:00
Oliver c3d7b92026 don't rely on stats code for new metrics if we can avoid it 2021-05-13 21:14:05 -04:00
Oliver 30f7226705 add prometheus metrics endpoint 2021-05-13 21:14:04 -04:00
tidwall 47a4e16c0b Add comment 2021-04-28 05:09:48 -07:00
tidwall d37ce505f5 Allow for all command types for roaming event
Prior to this commit roaming geofences only registered changes to
the SET command. Now it will work for SET/DEL/FSET, just like
non-roaming geofences.

To opt out of these events, explicitly choose which event you
would like to register when you create the geofences.

For example:

  NEARBY fleet FENCE COMMANDS set,del ROAM fleet * 5000

Will only fire off events from SET and DEL command.

Closes #597
2021-04-01 14:59:12 -07:00
tidwall 6b08f7fa9e Code cleanup
- Removed unused functions and variables
- Wrapped client formatted errors
- Updated deprecated packages
- Changed suggested code patterns
2021-03-31 08:13:44 -07:00
tidwall 40105e6b7e Include "distance" to output when user specifically requests
fixes #599
2021-03-30 11:49:01 -07:00
tidwall 72dfaaec63 Updated dependencies 2021-02-07 17:54:56 -07:00
tidwall 016f397124 Replace tinybtree 2021-01-25 14:34:40 -07:00
tidwall 9ce20331e4 Fixed fields being shuffled after AOFSHRINK 2020-11-09 14:45:40 -07:00
tidwall 1a7d8d6431 Added ENV var for 500 http errors 2020-11-07 05:22:11 -07:00
tidwall 9998e03f6f Optimization for non-cross geofence detection
This commit fixes a performance issue with the algorithm that
determines with geofences are potential candidates for
notifications following a SET operation.

Details

Prior to commit b471873 (10 commits ago) there was a bug where
the "cross" detection was not firing in all cases. This happened
because when looking for candidates for "cross" due to a SET
operation, only the geofences that overlapped the previous
position of the object and the geofences that overlapped the new
position where searched. But, in fac, all of the geofences that
overlapped the union rectangle of the old and new position should
have been searched.

That commit fixed the problem by searching a union rect of the
old and new positions. While this is an accurate solution, it
caused a slowdown on systems that have big/wild position changes
that might cross a huge number of geofences, even when those
geofences did not need actually need "cross" detection.

The fix

With this commit the geofences that have a "cross" detection
are stored in a seperated tree from those that do not. This
allows for a hybrid of the functionality prior and post b471873.

Fixes #583
2020-10-23 09:51:27 -07:00
tidwall 93e3a067b7 Truncate trailing zeros from AOF at startup
This commit addresses issue #230, where an AOF file will sometimes
not load due to the file being padded with trailing zeros. It's
uncertain what is causing this corruption, but it appears to be
coming from outside of the tile38-server process. I suspect it's
due to some block store layer in Kubernetes/Docker cloud
environments.

This fix allows for Tile38 to start up by discovering the trailing
zeros while loading the AOF and safely truncating the file as to
not include the zeros in the future.
2020-10-07 09:52:32 -07:00
tidwall b47187349b Fix "cross" detection on firing in some cases
Fixes #578
2020-09-22 16:19:49 -07:00
tidwall d819db5f8b Added MONITOR command
closes #571
2020-08-12 12:38:52 -07:00
tidwall 9093926135 Limit geohash precision to 12 2020-08-12 10:58:49 -07:00
Terra Brown 9202fd0206
s/64/12/ 2020-08-11 18:11:06 -04:00
tidwall f23b46cfc0 Hotfix for previous commit 2020-07-04 10:35:07 -07:00
tidwall f3622c3756 Fixed wrong order for fields with SCAN
This commit fixes an issue that happens when running SCAN on a
collection that has objects with fields, causing field values
to be mismatched with their respective keys.

This only occured with json output, and is a regression from #534.

Fixes #569
2020-07-04 10:12:06 -07:00
Lars Wilhelmsen 4fdbade7a4
Fix crash in fenceMatchRoam causing an index out of range panic 2020-06-18 15:43:18 +02:00
tidwall b7978f2cf4 Fix OUTPUT client command requiring authentication
This commit fixes an issue where the OUTPUT command requires
authentication when a server password has been set with
CONFIG SET requirepass. This was causing problems with clients
that use json responses, like the tile38-cli.

Fixes #564
2020-06-04 12:57:23 -07:00
Alex Roitman 25579a052c Fix a bug in WHEREIN -- 0 values would always match, incorrectly. 2020-04-12 16:06:10 -07:00
Alex Roitman 914f51de11 Fixes 2020-04-09 09:59:24 -07:00
Alex Roitman fe0216c42c Restore previous behavior where non-existing fields are treated as zero-value. 2020-04-09 09:54:47 -07:00
Alex Roitman f3cc365d24 Pre-allocate where and wherein arrays. 2020-04-09 09:36:12 -07:00
Mike Poindexter 2a4272c95f Improve kNN behavior
The current KNN implementation has two areas that can be improved:

- The current behavior is somewhat incorrect. When performing a kNN
query, the current code fetches k items from the index, and then sorts
these items according to Haversine distance. The problem with this
approach is that since the items fetched from the index are ordered by
a Euclidean metric, there is no guarantee that item k + 1 is not closer
than item k in great circle distance, and hence incorrect results can be
returned when closer items beyond k exist.

- The secondary sort is a performance killer. This requires buffering
all k items (again...they were already run through a priority queue in)
the index, and then a sort. Since the items are mostly sorted, and
Go's sort implementation is a quickSort this is the worst case for the
sort algorithm.

Both of these can be fixed by applying a proper distance metric in
the index nearby operation. In addition, this cleans up the code
considerably, removing a number of special cases that applied only
to NEARBY operations.

This change implements a geodetic distance metric that ensures that
the order from the index is correct, eliminating the need for the
secondary sort and special filtering cases in the ScanWriter code.
2020-04-07 20:10:58 -07:00
Alex Roitman 34cb2affdc Add clipby subcommand to INTERSECTS/WITHIN 2020-04-03 16:49:06 -07:00
tidwall 951fc58e02 Match geometry indexing to server config 2020-03-25 15:35:31 -07:00
Alex Roitman d5132a9eae Map field names to array indices in scanwriter, once per query. 2020-03-25 15:15:06 -07:00
Alex Roitman 9e7766b346 Array of values instead of map for whereins. 2020-03-25 15:14:51 -07:00
Alex Roitman 27c6980f82 Copy array and only loop if we need to pad. 2020-03-25 15:14:37 -07:00
Alex Roitman 91ef777771 Add benchmark for fieldMatch. 2020-03-25 14:49:21 -07:00
tidwall 5162ac5fd7 Stable sort roam notifications 2020-03-25 13:01:11 -07:00
tidwall a99df2892a Fixed false faraway notifications 2020-03-25 12:47:55 -07:00
tidwall ff48054d3d Fixed a missing faraway event for roaming geofences
This commit fixes a case where a roaming geofence will not fire
a "faraway" event when it's supposed to.

The fix required rewriting the nearby/faraway detection logic. It
is now much more accurate and takes overall less memory, but it's
also a little slower per operation because each object proximity
is checked twice per update. Once to compare the old object's
surrounding, and once to evaulated the new object. The two lists
are then used to generate accurate "nearby" and "faraway" results.
2020-03-22 11:54:56 -07:00
Josh Baker 0997f2e82b
Merge pull request #534 from rshura/optimize-scanwriter
Avoid sorting fields for each written object.
2020-03-22 06:33:33 -07:00
tidwall 65943e8f1a Fixed DEL geofence notifications missing "key" field
Closed #538
2020-03-21 17:48:31 -07:00
Mike Poindexter 84cabd77f9 Fix a concurrent write/read on the server conn map 2020-03-20 16:47:13 -07:00
Alex Roitman 5faccc3b4c Avoid sorting fields for each written object. 2020-03-03 13:39:43 -08:00
tidwall 66af8ab094 Fix tile38-cli from freezing with non-quoted geojson
This commit addresses an issue that began on 1.19 where the
deprecated tile38 native line protocol was removed in favor of
the more robust resp protocol. In turn the tile38 cli required
that all args are quoteless or quote escaped.

The commit ensures that the server returns the correct error
message and also loosens the strictness of the need for quoted
arguments in the tile38-cli.

fixes #513
2019-12-11 11:08:33 -07:00
tidwall 474ff810c0 Fixed panic on AOFSHRINK
closes #508
2019-11-17 07:25:25 -07:00
tidwall c084aeedc2 Code cleanup
This commit cleans up various Go code in the internal directory.
- Ensures comments on exported functions
- Changes all *Server receiver in all files to be "s", instead
  of mixed "c", "s", "server", etc.
- Silenced Go warnings for if/else with returns.
- Cleaned up import ordering.
2019-10-30 10:17:59 -07:00
Josh Baker 981d9ece42
Merge pull request #503 from JordanArmstrong/fix-stats-cpu
Fix fprintf type error in stats_cpu.go for non-linux/darwin builds
2019-10-29 15:29:38 -07:00
Josh Baker 90c2474e3d
Merge pull request #501 from JordanArmstrong/strict-jset-numbers
Strictly check if values passed to JSET are numbers
2019-10-29 15:28:40 -07:00
tidwall 23b016d192 Fix excessive memory usage for objects with TTLs
This commit fixes an issue where Tile38 was using lots of extra
memory to track objects that are marked to expire. This was
creating problems with applications that set big TTLs.

How it worked before:

Every collection had a unique hashmap that stores expiration
timestamps for every object in that collection. Along with
the hashmaps, there's also one big server-wide list that gets
appended every time a new SET+EX is performed.

From a background routine, this list is looped over at least
10 times per second and is randomly searched for potential
candidates that might need expiring. The routine then removes
those entries from the list and tests if the objects matching
the entries have actually expired. If so, these objects are
deleted them from the database. When at least 25% of
the 20 candidates are deleted the loop is immediately
continued, otherwise the loop backs off with a 100ms pause.

Why this was a problem.

The list grows one entry for every SET+EX. When TTLs are long,
like 24-hours or more, it would take at least that much time
before the entry is removed. So for databased that have objects
that use TTLs and are updated often this could lead to a very
large list.

How it was fixed.

The list was removed and the hashmap is now search randomly. This
required a new hashmap implementation, as the built-in Go map
does not provide an operation for randomly geting entries. The
chosen implementation is a robinhood-hash because it provides
open-addressing, which makes for simple random bucket selections.

Issue #502
2019-10-29 11:19:33 -07:00
Jordan Armstrong 324b3b06d9 Fix go vet error in stats_cpu.go for non-linux/darwin builds 2019-10-28 22:39:27 -03:00
Jordan Armstrong e0eca0d55e Replace isJsonNumber with version from gjson 2019-10-28 22:35:44 -03:00
Josh Baker df477bf3f4
Merge pull request #464 from rshura/area-expression
Add area expressions.
2019-10-28 13:45:10 -07:00
tidwall 10f7bfc445 Fix invalid pubsub format for json output
This commit fixes a bug that causes bad formatting for geofence
notifications when a client's output is set to JSON.

closes #499
2019-10-28 12:51:44 -07:00
Jordan Armstrong 41fb410e2e Strictly check if values to JSET are numbers
Fixes #493
2019-10-26 23:37:48 -03:00
tidwall 87185319b2 Fix JSET cancels expiry
issue #498
2019-10-08 09:45:46 -07:00
tidwall 639f6e2deb Replaced boxtree for rbang 2019-09-12 18:42:53 -07:00
tidwall 3d96b17258 GC pauses be gone 2019-09-04 12:47:30 -07:00
tidwall 2571ce5106 Minimize AOF buffer releases 2019-09-03 17:01:26 -07:00
tidwall 4bd6b4b838 Moved ReadMemStats into a background polling function
This will keep profile commands such as SERVER and STATS from
stopping the world.
2019-09-03 16:57:49 -07:00
tidwall e167e88e8f removed diag 2019-09-03 16:39:51 -07:00
tidwall b6884fce63 diagnostics 2019-09-03 16:35:42 -07:00
tidwall aedd972516 Fixed invalid count for nearby queries
closes #489
2019-08-22 15:53:37 -07:00
saltatory ec5a5342a2 Added RLock on connections map 2019-08-09 08:48:23 -07:00
tidwall f7888c1edf Fixed malformed json for chans command
Mentioned by ds2xor on Slack
2019-08-03 10:10:28 -07:00
Alex Roitman b9e2c67933 Fix lua pool pruning 2019-08-01 11:55:52 -07:00
Alex Roitman dd09ffbe13 Fix parser for multiple negations. 2019-06-14 10:02:26 -07:00
Alex Roitman 7c541949b1 Unnecessary code. 2019-06-13 14:12:42 -07:00
Alex Roitman eb214cb889 Better naming and comments. 2019-06-13 13:33:07 -07:00
Alex Roitman 2d83e18934 Add expression errors test. Make parser stricter. 2019-06-13 13:10:47 -07:00
Alex Roitman 0c3a5d02ca Fixes 2019-06-13 12:04:04 -07:00
Alex Roitman 3ded4e3a44 Refactor using method expressions, to avoid repeating code. 2019-06-13 10:56:33 -07:00
Alex Roitman 81f57ba6f6 Fix one case in expression/expression tests 2019-06-13 09:53:17 -07:00
Alex Roitman 496ace25d3 Refactoring/cleanup/fixes. 2019-06-11 17:13:33 -07:00
Alex Roitman d0ca579f5c Add area expressions.
Add parser.
Hook up to test command.
Add tests for expressions in test command.
2019-06-10 14:47:42 -07:00
Josh Baker aa73fcd17b Added unsubscribing from active channel
related #448
2019-06-04 13:49:32 -07:00
Melissa Baker 46f5d2d224 Fixed nearby json field results showing wrong data
issue #453
2019-05-28 12:17:59 -07:00
tidwall 3ae59274e3 Removed evio option 2019-04-26 11:50:49 -07:00
Alex Roitman cbf10b7472 Rework timeouts to allow prepending any command with the TIMEOUT seconds 2019-04-25 14:15:53 -07:00
Alex Roitman 49111a4dce Add timeouts for lua scripts 2019-04-24 17:00:52 -07:00
Alex Roitman 31525487c3 Add timeout tests. Fix a bug. 2019-04-24 13:20:57 -07:00
Alex Roitman e514a0287f Add timeout subcommand to scan/search commands.
Use per-query timeout for those commands, if it was given.
2019-04-24 12:02:39 -07:00
tidwall 0aecef6a5c Added TIMEOUT command 2019-04-24 05:09:41 -07:00
tidwall bb33467da1 Fix crash with fence ROAM
closes #440
2019-04-23 11:16:55 -07:00
tidwall 8c69fbff31 Fixed clients blocking while webook sending
Closes #428
2019-03-16 15:43:33 -07:00
tidwall 762607dc0a Added geofence delivery counter 2019-03-14 11:23:23 -07:00
tidwall 5333fab870 Recycle aof buffer 2019-03-10 10:48:14 -07:00
tidwall fc3e8b4359 Fix nearby with match query invalid results
closes #421
2019-03-01 06:55:26 -07:00
Steven Wolfe fd20190bff Verify hook names match for processing 2019-02-22 15:58:13 -07:00
Parham Alvani e0ddd5e6a6 Adds Server command for lua scripts. 2019-02-17 13:28:20 +03:30
tidwall fb7259b10b Changed clip errors and json result type 2019-02-12 05:33:20 -07:00
tidwall 62f44ed055 Merge branch 'master' of https://github.com/tidwall/tile38 2019-02-12 05:06:10 -07:00
Josh Baker 6fffa75b58
Merge pull request #409 from rshura/test_cmd
Test command
2019-02-12 05:05:50 -07:00
tidwall 14b616b150 More test fields 2019-02-11 15:42:35 -07:00
tidwall 92c1ce8ef9 Update tinybtree dep 2019-02-11 13:39:29 -07:00
Alex Roitman c849ab19ac Implement test command 2019-02-08 13:57:29 -08:00
Alex Roitman f7ba43360e Typo in error messages 2019-02-08 13:56:07 -08:00
Steven Wolfe 79c902efbf Adding more replication data to INFO response 2019-01-18 14:51:20 -07:00
Steven Wolfe ffd8ca915c Testing for valid INFO and CLIENT Json output 2019-01-15 11:08:19 -07:00
Steven Wolfe 0740ca13cd JSON Output for INFO and CLIENT 2019-01-14 15:29:29 -07:00
Steven Wolfe 4392fb44bd Fixing KEYS command and tests 2019-01-14 11:06:12 -07:00
Steven Wolfe 4ad2e920f5 Merge branch 'sl/requirevalid' of https://github.com/stevelacy/tile38 into stevelacy-sl/requirevalid 2019-01-10 12:49:48 -07:00
Steve Lacy 1a8993feaf
Add RequireValid env variable and parse options 2019-01-10 09:36:58 -07:00
Steven Wolfe eb2ff76f86 Sorting > 1 and docs 2019-01-09 10:35:50 -07:00
Steven Wolfe a5fbcf5f15 Ensuring channel publish order 2019-01-09 00:23:53 -07:00
tidwall ed31f2bd4c Merge branch 'lua/distance_to' of https://github.com/rshura/tile38 into rshura-lua/distance_to 2019-01-06 10:29:59 -07:00
tidwall d7d0baa954 Fix roam fence missing 2019-01-06 10:23:57 -07:00
Alex Roitman 5e215c247d Add distance_to function to the tile38 namespace in lua.
distance_to(lat_a, lon_a, lat_b, lon_b) returns distance in meters between points A and B.
2019-01-04 14:57:00 -08:00
Alex Roitman e1c82e17f7 Refactor to be more consistent with redis responses. 2018-12-28 14:46:54 -08:00
Alex Roitman 01a7dda2a1 Add RENAME and RENAMENX commands. 2018-12-27 17:15:53 -08:00
tidwall 1c972ad578 Fix massinsert dev call 2018-12-03 15:35:32 -07:00
tidwall f795567f9c Merge branch 'master' into Ext-server-stats 2018-11-29 15:22:48 -08:00
tidwall 6b2fc1f37f Fix multiple matches on different key hooks 2018-11-29 15:15:26 -08:00
Steven Wolfe 8619d35a5b Adding ext to SERVER command for more info 2018-11-26 14:19:08 -07:00
tidwall 8b29e98359 Optimized spatial index for fences 2018-11-23 18:15:14 -07:00
tidwall f2c217c216 Refactor and comment command details 2018-11-23 15:53:33 -07:00
tidwall 1197cca40c Avoid extra broadcasts 2018-11-23 15:38:49 -07:00
tidwall 8906d8e65a Refactor and consolidate variables 2018-11-23 02:14:26 -07:00
tidwall 37531f9350 Removed unused atomics 2018-11-23 01:39:04 -07:00
tidwall 58421ef806 Removed unused code 2018-11-23 01:24:35 -07:00
tidwall 669823950a Fix script load issue 2018-11-20 10:25:48 -08:00
tidwall 737561fa8b Fix version not being set at build, close #386 2018-11-15 13:45:55 -07:00
tidwall a0f3b0049f Fix aof loading problem 2018-11-15 09:15:39 -07:00
Josh Baker 37afa7e92b
Merge pull request #385 from stevelacy/sl/server_version
Add server command for `version` - closes #384
2018-11-15 04:18:28 -07:00
tidwall 4c3ac62191 Hotfix replica sync needs flushing 2018-11-13 12:04:16 -07:00
Steve Lacy 5c6c280559
Add server command for `version` - closes #384 2018-11-13 10:24:15 -07:00
tidwall 6616b86eda Default numloops to number goprocs 2018-11-11 09:29:07 -07:00
tidwall 7cc9154eb8 Correct threads equals cpu 2018-11-11 09:05:26 -07:00
tidwall d065b979da net prewrite optimization 2018-11-11 09:05:26 -07:00
tidwall 464c193d73 Prewrite optimization flag 2018-11-11 09:05:26 -07:00
tidwall 3bf91077d9 Hang on to lower command 2018-11-11 09:05:26 -07:00
tidwall 161c6faff9 Added evio flag 2018-11-11 09:03:47 -07:00
tidwall 933f243c6c Code cleanup 2018-11-11 09:03:47 -07:00