apitech/loki - loki - Gitea: Git with a cup of tea

Commit Graph

Author	SHA1	Message	Date
Owen Diehl	ad803af9dc	Wal/recover corruption (#3117 ) * simplifies recovery code and continues after errors * does not exit on wal recovery failure * log line parity * less scary error msgs * more descriptive errors * Update pkg/ingester/ingester.go Co-authored-by: Ed Welch <ed@oqqer.com> * Update pkg/ingester/ingester.go Co-authored-by: Ed Welch <ed@oqqer.com> Co-authored-by: Ed Welch <ed@oqqer.com>	4 years ago
Cyril Tovena	7682f133b6	Improve JSON parser and add labels parser hints. (#3080 ) * Improve JSON parser and add labels parser hints. This is a rework of the json parser, that now uses iteration over reflection allowing to avoid allocations. I've also implemented a way to guess and hints what labels needs to be parsed when doing metric queries, this means we can extract only the few labels required from the log line. benchcmp ``` ❯ benchcmp before.txt after.txt5 benchmark old ns/op new ns/op delta Benchmark_Parser/json/no_labels_hints-16 9889 3281 -66.82% Benchmark_Parser/logfmt/no_labels_hints-16 1624 1671 +2.89% Benchmark_Parser/logfmt/labels_hints-16 1601 790 -50.66% benchmark old allocs new allocs delta Benchmark_Parser/json/no_labels_hints-16 139 56 -59.71% Benchmark_Parser/logfmt/no_labels_hints-16 31 31 +0.00% Benchmark_Parser/logfmt/labels_hints-16 31 5 -83.87% benchmark old bytes new bytes delta Benchmark_Parser/json/no_labels_hints-16 3671 912 -75.16% Benchmark_Parser/logfmt/no_labels_hints-16 464 464 +0.00% Benchmark_Parser/logfmt/labels_hints-16 464 144 -68.97% ``` I've experienced a 2x to 4x improvement in my cluster. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * typos. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Review feedback. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	4 years ago
Owen Diehl	58987553a1	Disables the stream limiter until wal has recovered (#3114 ) * disables the stream limiter until wal has recovered * limiter disable/enable methods	4 years ago
Owen Diehl	b236f2ee27	adds the ability to read chunkFormatV3 while writing v2 (#2984 ) (cherry picked from commit `a8af1c0c84`)	4 years ago
Michel Hollands	48b65834a0	Make server publicly available (#3087 ) Signed-off-by: Michel Hollands <michel.hollands@grafana.com>	5 years ago
Michel Hollands	e9001d8731	Make the moduleManager field public (#3077 ) Signed-off-by: Michel Hollands <michel.hollands@grafana.com>	5 years ago
Owen Diehl	35a627df49	adds the ability to read chunkFormatV3 while writing v2 (#2984 ) (cherry picked from commit `a8af1c0c84`)	5 years ago
Cyril Tovena	fbf17904f7	Fixes iota unexpected behaviour with bytes for chunk encoding. (#2945 ) * Fixes iota unexpected behaviour with bytes for chunk encoding. This was introducing incompatibilities since iota with bytes starts at 2. Seems like a weird conversion between iota int and byte. This can currenlty corrupt chunks, and previous chunks are not decodable. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * With iota correclty this time. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> (cherry picked from commit `fa7ea9cc2d`)	5 years ago
Edward Welch	8cd276d4ac	v3 of chunk encoding serializes all block data (#2934 ) (cherry picked from commit `663c872a4c`)	5 years ago
Cyril Tovena	6d85c7c212	Add absent_over_time (#3053 ) * Add absent_over_time This adds absent_over_time which is super useful when alerting, it can be combine with any logql selector including parsers and filter to returns a timeseries with 1 when logs go missing. Fixes #2893 Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Adds comment. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Cyril Tovena	3f0800dc44	Add function label_replace. (#3047 ) You might be wondering why do we need label_replace when we already have label_format. Well for two reasons: 1. It's nice to have something similar to Prometheus, so users building metric queries will feel right at home. 2. Interestingly, they are quite different in how they execute. label_format executes for every log line, however label_replace execute for every sample. Since you aggregate log lines into sample then label_replace is way more efficient at scale. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Cyril Tovena	4e7a123c77	Support rate for unwrap expressions. (#3048 ) This allows to bytes throughput if there were printed in a log line. Examples : ```logql sum by (device) (rate({app="network-dev"} \|= "192.16.8.1" \| json \| unwrap recv_bytes [1m])) ``` Closes #3044 Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Cyril Tovena	2406ea72bc	Allows by/without to be empty and available for max/min_over_time (#3030 ) * Allows by/without to be empty and available for max/min_over_time operations. This essentially allows to aggregate over all dimensions when using by () while without() is a noop. This also makes it possible for max and min range vector aggregation to use grouping, which is simpler than doing max by (foo) max_over_time(...) Examples: - `max_over_time(...) by ()` gives your the max over time accross of all series. - `min_over_time(...) by (namespace)` gives you the min over time per namespace. - `max_over_time(...) without (namespace)` gives you the max over time removing the namespace dimension. PS: I've also refactored a bit how we optimized grouping to make it more clear. Fixes #2884 Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * nit doc. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
AdamKorcz	a3ae2f08f9	Added fuzzer (#3031 )	5 years ago
Reinhard Nägele	1bb4301f2e	Fix multiline docs (#3035 ) Signed-off-by: Reinhard Nägele <unguiculus@gmail.com>	5 years ago
Karsten Jeschkies	afc1e53df9	Collapse multiline logs based on a start line. (#3024 ) * Collapse multiline logs based on a start line. Summary: This is a very simple approach based on #1380 to provide multiline or block log entries in promtail. A `multiline` stage is added to pipelines. This stages matches a start line. Once a start line is matched all following lines are appended to an entry and not passed on to downstream stages. Once a new start line is matched the former block of multilines is sent. If now new line arrives withing `max_wait_time` the block is flushed to the next stage and a new block is started. * Test multiline stage process. * Format code. * Flush multiline block after `max_lines`. * Capture internal state of the stage. * Process different multiline streams in parallel. * Start documenting multiline stage. * Give an example configuration for `multiline` stage. * Make linter happy. * Pass through entries until first start line. * Update pkg/logentry/stages/multiline.go	5 years ago
Cyril Tovena	c98dfa0da1	Fixes a swallowed context err in the batch storage. (#3025 ) Fixes #2999 There's still a potential cases where, the context is done (and so the Err is not nil) and we're finished with iterating over the list of batches. I didn't introduce the error here because I think I would prefer a success in this case. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Cyril Tovena	c4faa579e3	Asynchronous Promtail stages (#2996 ) * Introducing go pipelines to promtail. Based off @jeschkies idea. WIP Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Fixes all tests. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * More tests and code cleanup. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Wip breaking things. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Fixing tests, adding Stop to the interface. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Fixes all tests. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Fixes more test. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Moar fixes for tests. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Close correctly client before reading. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Use defer. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Add some comments. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Fixes lint issues Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Attempt to fix journald without linux. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Fixes journald json test. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Add missing stop in the filetarget. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Update pkg/logentry/stages/match_test.go Co-authored-by: Karsten Jeschkies <k@jeschkies.xyz> * First set of feeback review fixes. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Use newEntry as suggested by Karsten. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Fix tests. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * lint. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> Co-authored-by: Karsten Jeschkies <k@jeschkies.xyz>	5 years ago
Owen Diehl	05f3d6e0da	Updates chunkenc doc for V3 (#2956 )	5 years ago
Cyril Tovena	315972409b	Add more information about why a tailer would stop. (#3015 ) Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Owen Diehl	854930e661	headblock checkpointing up to v3 (#3013 )	5 years ago
Cyril Tovena	dcbfecf9e5	Fixes the ruler storage with the boltdb store. (#3008 ) This ensure we have read-only local index, and also query chunk store from ingesters. Found this bug while deploying a new cluster using the ruler and boltdb store. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Owen Diehl	a8af1c0c84	adds the ability to read chunkFormatV3 while writing v2 (#2984 )	5 years ago
Kaviraj	410b9d9adc	fix(docker-driver): Propagate promtail's `client.Stop` properly (#2898 ) * impr(Makefile): add support to take docker-driver image via ENV * impr(promtail/client): Add cancelling mechanism for batchSend stores `context` and its `cancel` in client and uses it to make upstream calls that need `ctx`. Also properly cancel's all the upstream calls that using `ctx` when client's Stop is called. * Add `StopNow()` method to stop the client without retries. Use `StopNow()` from the docker-driver * Add test for StopNow * reverting back the client.Timeout	5 years ago
Owen Diehl	8c1fe88409	Introduces per stream chunks mutex (#3000 ) * marshalable chunks * wal record types custom serialization * proto types for wal checkpoints * byteswith output unaffected by buffer * wal & record pool ifcs * wal record can hold entries from multiple series * entry pool * ingester uses noopWal * removes duplicate argument passing in ingester code. adds ingester config validation & derives chunk encoding. * segment writing * [WIP] wal recovery from segments * replay uses sync.Maps & preserves WAL fingerprints * in memory wal recovery * wal segment recovery * ingester metrics struct * wal replay locks streamsMtx in instances, adds checkpoint codec * ingester metrics * checkpointer * WAL checkpoint writer * checkpointwriter can write multiple checkpoints * reorgs checkpointing * wires up checkpointwriter to wal * ingester SeriesIter impl * wires up ingesterRecoverer to consume checkpoints * generic recovery fn * generic recovery fn * recover from both wal types * cleans up old tmp checkpoints & allows aborting in flight checkpoints * wires up wal checkpointing * more granular wal logging * fixes off by 1 wal truncation & removes double logging * adds userID to wal records correctly * wire chunk encoding tests * more granular wal metrics * checkpoint encoding test * ignores debug bins * segment replay ignores out of orders * fixes bug between WAL reading []byte validity and proto unmarshalling refs * conf validations, removes comments * flush on shutdown config * POST /ingester/shutdown * renames flush on shutdown * wal & checkpoint use same segment size * writes entries to wal regardless of tailers * makes wal checkpoing duration default to 5m * recovery metrics * encodes headchunks separately for wal purposes * merge upstream * linting * adds chunkMtx to ingester streams * all flush locking locks streams -> chunks in order to prevent deadlocks * wal integration tests * addresses pr feedback uses entry pool in stream push/tailer removes unnecessary pool interaction checkpointbytes comment fillchunk helper, record resetting in tests via pool redundant comment defers wg done in recovery s/num/count/ checkpoint wal uses a logger encodeWithTypeHeader now creates its own []byte removes pool from decodeEntries wal stop can error * cleans up merge * lint * inline functions to simplify unlocking	5 years ago
Owen Diehl	4d9865acd4	Adds WAL support (experimental) (#2981 ) * marshalable chunks * wal record types custom serialization * proto types for wal checkpoints * byteswith output unaffected by buffer * wal & record pool ifcs * wal record can hold entries from multiple series * entry pool * ingester uses noopWal * removes duplicate argument passing in ingester code. adds ingester config validation & derives chunk encoding. * segment writing * [WIP] wal recovery from segments * replay uses sync.Maps & preserves WAL fingerprints * in memory wal recovery * wal segment recovery * ingester metrics struct * wal replay locks streamsMtx in instances, adds checkpoint codec * ingester metrics * checkpointer * WAL checkpoint writer * checkpointwriter can write multiple checkpoints * reorgs checkpointing * wires up checkpointwriter to wal * ingester SeriesIter impl * wires up ingesterRecoverer to consume checkpoints * generic recovery fn * generic recovery fn * recover from both wal types * cleans up old tmp checkpoints & allows aborting in flight checkpoints * wires up wal checkpointing * more granular wal logging * fixes off by 1 wal truncation & removes double logging * adds userID to wal records correctly * wire chunk encoding tests * more granular wal metrics * checkpoint encoding test * ignores debug bins * segment replay ignores out of orders * fixes bug between WAL reading []byte validity and proto unmarshalling refs * conf validations, removes comments * flush on shutdown config * POST /ingester/shutdown * renames flush on shutdown * wal & checkpoint use same segment size * writes entries to wal regardless of tailers * makes wal checkpoing duration default to 5m * recovery metrics * encodes headchunks separately for wal purposes * merge upstream * linting * addresses pr feedback uses entry pool in stream push/tailer removes unnecessary pool interaction checkpointbytes comment fillchunk helper, record resetting in tests via pool redundant comment defers wg done in recovery s/num/count/ checkpoint wal uses a logger encodeWithTypeHeader now creates its own []byte removes pool from decodeEntries wal stop can error * prevent shared access bug with tailers and entry pool * removes stream push entry pool optimization	5 years ago
Sandeep Sukhani	ae9c4b82ec	fix table deletion in table client for boltdb-shipper (#2960 )	5 years ago
Ed Welch	9ba5753e25	Loki: handle faults when opening boltdb files (#2988 ) * we also need to catch faults not just panics when opening boltdb files * fix lint * i'll get it one of these times * fix typo	5 years ago
Cyril Tovena	ec725db9d0	Add ProcessString to Pipeline. (#2972 ) * Add ProcessString to Pipeline. Most of the time, you have a buffer that you want to test again a log pipeline, when it passed you usually copy via string(buff). In some cases like headchunk and tailer you already have an allocated immutable string, since pipeline never mutate the line passed as parameter, we can create ProcessString pipeline method that will avoid re-allocating the line. benchmp: ``` ❯ benchcmp before.txt after.txt benchmark old ns/op new ns/op delta BenchmarkHeadBlockIterator/Size_100000-16 20264242 16112591 -20.49% BenchmarkHeadBlockIterator/Size_50000-16 10186969 7905259 -22.40% BenchmarkHeadBlockIterator/Size_15000-16 3229052 2202770 -31.78% BenchmarkHeadBlockIterator/Size_10000-16 1916537 1392355 -27.35% BenchmarkHeadBlockSampleIterator/Size_100000-16 18364773 16106425 -12.30% BenchmarkHeadBlockSampleIterator/Size_50000-16 8988422 7730226 -14.00% BenchmarkHeadBlockSampleIterator/Size_15000-16 2788746 2306161 -17.30% BenchmarkHeadBlockSampleIterator/Size_10000-16 1773766 1488861 -16.06% benchmark old allocs new allocs delta BenchmarkHeadBlockIterator/Size_100000-16 200039 39 -99.98% BenchmarkHeadBlockIterator/Size_50000-16 100036 36 -99.96% BenchmarkHeadBlockIterator/Size_15000-16 30031 31 -99.90% BenchmarkHeadBlockIterator/Size_10000-16 20029 29 -99.86% BenchmarkHeadBlockSampleIterator/Size_100000-16 100040 40 -99.96% BenchmarkHeadBlockSampleIterator/Size_50000-16 50036 36 -99.93% BenchmarkHeadBlockSampleIterator/Size_15000-16 15031 31 -99.79% BenchmarkHeadBlockSampleIterator/Size_10000-16 10029 29 -99.71% benchmark old bytes new bytes delta BenchmarkHeadBlockIterator/Size_100000-16 27604042 21203941 -23.19% BenchmarkHeadBlockIterator/Size_50000-16 13860654 10660652 -23.09% BenchmarkHeadBlockIterator/Size_15000-16 4231360 3271363 -22.69% BenchmarkHeadBlockIterator/Size_10000-16 2633436 1993420 -24.30% BenchmarkHeadBlockSampleIterator/Size_100000-16 17799137 14598973 -17.98% BenchmarkHeadBlockSampleIterator/Size_50000-16 7433260 5833137 -21.53% BenchmarkHeadBlockSampleIterator/Size_15000-16 2258099 1778096 -21.26% BenchmarkHeadBlockSampleIterator/Size_10000-16 1393600 1073605 -22.96% ``` Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * add some precision about why Sum64 is not a concern. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Update pkg/chunkenc/memchunk.go Co-authored-by: Owen Diehl <ow.diehl@gmail.com> Co-authored-by: Owen Diehl <ow.diehl@gmail.com>	5 years ago
Cyril Tovena	06a89ea0ee	Fixes vector grouping injection. (#2975 ) * Fixes vector grouping injection. During an improvement I introduced a bug where we would pre-aggregate some range vector aggregations, although the result is not the same. We can only inject vector grouping in range vector if the operation is associative through grouping. This is not the case for max/min/stddev/stddvar/quantile/avg_over_time. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * got linted. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Ed Welch	6d442673b7	adding a function to safely open boltdb files and recover from any panics thrown by boltdb when opening a file. (#2983 )	5 years ago
Cyril Tovena	9807f7c43e	Avoid parsing labels when tailer is sending from a stream. (#2973 ) * Avoid parsing labels when tailer is sending from a stream. This also include some code cleanup. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * I got linted. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Cyril Tovena	8ea5fd173f	Implement io.WriteTo by chunks. (#2962 ) This allows to totally defer the buffer creation to the user. `Marshal(w io.Writer) error` Now don't even need to create a buffer. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Cyril Tovena	a62f81f1cf	Improve tailer matcher function. (#2959 ) Don't create a map on every new push of a stream or appearance of a new stream. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Cyril Tovena	4da505b413	Fixes the Stringer of the byte label operator. (#2946 ) This would cause the operator to become a numeric one once stringified. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Karsten Jeschkies	29be929a8c	Test label filter for bytes. (#2941 ) * Test label filter for bytes. Summary: - Adds a new test to `ast_test` for filtering by bytes.\| - Test another case than `B`. * Test all combinations of byte sizes.	5 years ago
Sandeep Sukhani	cd723c0479	Boltdb shipper query readiness (#2911 ) * add support for downloading index on startup in boltdb-shipper for query readiness * tests * pass context to load tables function to be able to cancel the operation * add missing call to tablemanager.stop when loading of tables fail	5 years ago
Cyril Tovena	61ba02e2c4	Limit series for metric queries. (#2903 ) * Work in progress limit series. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * First draft of the series limiter. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Add limiter to query-frontend. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Add documentation. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Fix build Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * fmted Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * lint and build fix Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Update docs/sources/configuration/_index.md Co-authored-by: achatterjee-grafana <70489351+achatterjee-grafana@users.noreply.github.com> * Review feedback. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Reduce // for to fix flaky test. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> Co-authored-by: achatterjee-grafana <70489351+achatterjee-grafana@users.noreply.github.com>	5 years ago
Cyril Tovena	fa7ea9cc2d	Fixes iota unexpected behaviour with bytes for chunk encoding. (#2945 ) * Fixes iota unexpected behaviour with bytes for chunk encoding. This was introducing incompatibilities since iota with bytes starts at 2. Seems like a weird conversion between iota int and byte. This can currenlty corrupt chunks, and previous chunks are not decodable. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * With iota correclty this time. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Cyril Tovena	14a5fda15b	Only append tailed entries if needed. (#2931 ) * Only append tailed entries if needed. This avoid to doing allocations on slices when there is no tailers. We could potentially pooled the slice too but the tailing code is using the stream in an async fashion so we can't put it back. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * cleanup Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Owen Diehl	663c872a4c	v3 of chunk encoding serializes all block data (#2934 )	5 years ago
Cyril Tovena	58f8117618	Skip journald bad message. (#2928 ) Unfortunatelly I can't show the message itself without forking the library we are using. Fixes #2812. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Cyril Tovena	e1ab2cc509	Cleanup labels parsing. (#2929 ) * Cleanup labels parsing. Make sure we only use one type of labels parsing.This just reduce the amount of code we have to maintain. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Fixes tests. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Fixes a test. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Cyril Tovena	18e29236f4	Fixes all in one grpc registrations. (#2930 ) We can only register push server once, in all in one, we want to register only ingester. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Cyril Tovena	049c7acc12	Improve logql parser allocations. (#2927 ) * Improve logql parser allocations. Also moving labels parsing into logql package, and I'm not shy to say that promql is faster for this. So we're using promql for parser, it's faster because their lexer is more sophisticated than ours. Result are impressive. ``` ❯ benchcmp before.txt after.txt benchmark old ns/op new ns/op delta Benchmark_ParseLabels-16 24394 8745 -64.15% benchmark old allocs new allocs delta Benchmark_ParseLabels-16 156 20 -87.18% benchmark old bytes new bytes delta Benchmark_ParseLabels-16 24976 2099 -91.60% ``` Combined with #2926, I think we should be able to get rid of `util.ToClientLabels`. But that's for another PR. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Fixes a test. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Cyril Tovena	2303f1bfa1	Cache label strings in ingester to improve memory usage. (#2926 ) * Add benchmark for base stats. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * Cache labels string in ingester to improve memory usage. This reduce allocations by a lot, but assume that labels comes in sorted which is not ensure in the distributors. The distributors will take a performance hit, but that's easier to scale or improve later.(added some todos) see benchmark: ``` ❯ benchcmp before.txt after.txt benchmark old ns/op new ns/op delta Benchmark_PushInstance-16 43505 4950 -88.62% benchmark old allocs new allocs delta Benchmark_PushInstance-16 240 12 -95.00% benchmark old bytes new bytes delta Benchmark_PushInstance-16 42568 1787 -95.80% ``` Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com> * 🤦 wrong hashing. Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago
Tom Manning	a5cc650d33	Distributor registers logproto.Pusher service to receive logs via GRPC (#2922 )	5 years ago
Chance Zibolski	f0d4adc53d	promtail: Add support for using syslog message timestamp (#2914 ) Currently promtail sets the timestamp of incoming syslog messages to the time it was received by promtail. In some cases, it is preferable to use the source timestamp instead. This adds a new `use_message_timestamp` option to the syslog target config, which allows users to opt-in to the behavior of using the timestamp on the message, if one exists.	5 years ago
Owen Diehl	bdf8dca5b1	loki delete namespace ruler api (#2918 )	5 years ago
Cyril Tovena	6e79cee85d	Fixes race condition in tailer since logql v2. (#2917 ) Signed-off-by: Cyril Tovena <cyril.tovena@gmail.com>	5 years ago

1 2 3 4 5 ...

835 Commits (e8cce09de29dc58c38b333e72dfe11613d49fbb3)