* Update go to 1.19, set min version to 1.18
Signed-off-by: Julien Pivotto <roidelapluie@o11y.eu>
* Update golangci-lint
Signed-off-by: Julien Pivotto <roidelapluie@o11y.eu>
Signed-off-by: Julien Pivotto <roidelapluie@o11y.eu>
Metadata was added recently but doesn't seem to be used much, at least as far as I could identify.
Yet it's part of memSeries struct and so even when empty takes 48 bytes,
which is a lot given that without it memSeries requires 224 bytes.
This change turns it into a pointer on the struct, that get set only when metadata is actually set of given series.
Signed-off-by: Łukasz Mierzwa <l.mierzwa@gmail.com>
Signed-off-by: Łukasz Mierzwa <l.mierzwa@gmail.com>
We're loading the time window once per each Append() call, plus once in
the Commit(). While not extremely expensive, atomic operations are also
not cheap.
Additionally, it makes sense to keep the window consistent for a single
append.
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
OOOHeadIndexReader was using the headIndexReader PostingsForMatchers()
and LabelValues() implementation which lead to a very subtle bug that
led to wrong query results.
headIndexReader LabelValues() implementation checks if the query
timerange overlaps with the head maxt and mint and if it doesnt it
returns an empty list of values. Since this code was also used by the
ooo head it led to wrong results that we were not able to see in tests
because our queries where always from MinInt64 to MaxInt64. This commit
also adds a new test that performs multiple time range queries to make
sure this never happens again.
Signed-off-by: Jesus Vazquez <jesus.vazquez@grafana.com>
* Add histogram validation
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Correct negative offset validation
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Address review comments
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Validation benchmark
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Add more checks
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Attempt to fix tests
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Fix stuff
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* tsdb/record: Extract functions to encode and decode labels
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
* tsdb: make use of Encode/Decode Labels
Simplify the code by re-using routines from tsdb/record.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
* Append metadata to the WAL
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Remove extra whitespace; Reword some docstrings and comments
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Use RLock() for hasNewMetadata check
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Use single byte for metric type in RefMetadata
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Update proposed WAL format for single-byte type metadata
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Implementa MetadataAppender interface for the Agent
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Address first round of review comments
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Amend description of metadata in wal.md
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Correct key used to retrieve metadata from cache
When we're setting metadata entries in the scrapeCace, we're using the
p.Help(), p.Unit(), p.Type() helpers, which retrieve the series name and
use it as the cache key. When checking for cache entries though, we used
p.Series() as the key, which included the metric name _with_ its labels.
That meant that we were never actually hitting the cache. We're fixing
this by utiling the __name__ internal label for correctly getting the
cache entries after they've been set by setHelp(), setType() or
setUnit().
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Put feature behind a feature flag
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Fix AppendMetadata docstring
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Reorder WAL format document
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Change error message of AppendMetadata; Fix access of s.meta in AppendMetadata
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Reuse temporary buffer in Metadata encoder
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Only keep latest metadata for each refID during checkpointing
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Fix test that's referencing decoding metadata
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Avoid creating metadata block if no new metadata are present
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Add tests for corrupt metadata block and relevant record type
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Fix CR comments
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Extract logic about changing metadata in an anonymous function
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Implement new proposed WAL format and amend relevant tests
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Use 'const' for metadata field names
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Apply metadata to head memSeries in Commit, not in AppendMetadata
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Add docstring and rename extracted helper in scrape.go
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Add tests for tsdb-related cases
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Fix linter issues vol1
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Fix linter issues vol2
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Fix Windows test by closing WAL reader files
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Use switch instead of two if statements in metadata decoding
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Fix review comments around TestMetadata* tests
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Add code for replaying WAL; test correctness of in-memory data after a replay
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Remove scrape-loop related code from PR
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Address first round of comments
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Simplify tests by sorting slices before comparison
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Fix test to use separate transactions
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Empty out buffer and record slices after encoding latest metadata
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Fix linting issue
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Update calculation for DroppedMetadata metric
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Rename MetadataAppender interface and AppendMetadata method to MetadataUpdater/UpdateMetadata
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Reuse buffer when encoding latest metadata for each series
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Fix review comments; Check all returned error values using two helpers
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Simplify use of helpers
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Satisfy linter
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Add exemplar record case
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* recordType() -> Type.String()
Signed-off-by: Levi Harrison <git@leviharrison.dev>
Previously, the maxTime wasn't updated properly in case of a recoding
happening.
My apologies for reformatting many lines for line length. During the
bug hunt, I tried to make things more readable in a reasonably wide
editor window.
Signed-off-by: beorn7 <beorn@grafana.com>
Previously, the maxTime wasn't updated properly in case of a recoding
happening.
My apologies for reformatting many lines for line length. During the
bug hunt, I tried to make things more readable in a reasonably wide
editor window.
Signed-off-by: beorn7 <beorn@grafana.com>
Relates to @bboreham optimization in https://github.com/prometheus/prometheus/pull/10859
Bryan did reduce the sleep time improving the deltas on the benchmark by
quite a lot. However I've been working on a similar implementation for
out of order and I noticed that we actually get into this method
thousands of times.
@ywwg had the brilliant idea of not always sleeping before the select
but actually make it a case in the select so we only sleep if we need
to.
The benchmark deltas are amazing
```
❯ benchstat old_implementation.txt new_implementation_using_time_after.txt
name old time/op new time/op delta
LoadWAL/batches=10,seriesPerBatch=100,samplesPerSeries=7200,exemplarsPerSeries=0,mmappedChunkT=0-8 521ms ±25% 253ms ± 6% -51.47% (p=0.008 n=5+5)
LoadWAL/batches=10,seriesPerBatch=100,samplesPerSeries=7200,exemplarsPerSeries=36,mmappedChunkT=0-8 773ms ± 3% 369ms ±31% -52.23% (p=0.008 n=5+5)
LoadWAL/batches=10,seriesPerBatch=100,samplesPerSeries=7200,exemplarsPerSeries=72,mmappedChunkT=0-8 592ms ±28% 297ms ±28% -49.80% (p=0.008 n=5+5)
LoadWAL/batches=10,seriesPerBatch=100,samplesPerSeries=7200,exemplarsPerSeries=360,mmappedChunkT=0-8 547ms ± 2% 999ms ±187% ~ (p=0.690 n=5+5)
LoadWAL/batches=10,seriesPerBatch=10000,samplesPerSeries=50,exemplarsPerSeries=0,mmappedChunkT=0-8 11.3s ± 4% 1.3s ±44% -88.48% (p=0.008 n=5+5)
LoadWAL/batches=10,seriesPerBatch=10000,samplesPerSeries=50,exemplarsPerSeries=2,mmappedChunkT=0-8 11.1s ± 1% 1.2s ±20% -89.08% (p=0.008 n=5+5)
LoadWAL/batches=10,seriesPerBatch=1000,samplesPerSeries=480,exemplarsPerSeries=0,mmappedChunkT=0-8 1.24s ± 3% 0.18s ± 7% -85.76% (p=0.008 n=5+5)
LoadWAL/batches=10,seriesPerBatch=1000,samplesPerSeries=480,exemplarsPerSeries=2,mmappedChunkT=0-8 1.24s ± 2% 0.18s ± 5% -85.24% (p=0.008 n=5+5)
LoadWAL/batches=10,seriesPerBatch=1000,samplesPerSeries=480,exemplarsPerSeries=5,mmappedChunkT=0-8 1.23s ± 5% 0.27s ±33% -77.73% (p=0.008 n=5+5)
LoadWAL/batches=10,seriesPerBatch=1000,samplesPerSeries=480,exemplarsPerSeries=24,mmappedChunkT=0-8 1.28s ± 1% 0.36s ± 7% -71.51% (p=0.008 n=5+5)
LoadWAL/batches=100,seriesPerBatch=1000,samplesPerSeries=480,exemplarsPerSeries=0,mmappedChunkT=3800-8 12.1s ± 1% 3.1s ± 6% -74.33% (p=0.008 n=5+5)
LoadWAL/batches=100,seriesPerBatch=1000,samplesPerSeries=480,exemplarsPerSeries=2,mmappedChunkT=3800-8 12.1s ± 1% 3.4s ± 4% -71.94% (p=0.008 n=5+5)
LoadWAL/batches=100,seriesPerBatch=1000,samplesPerSeries=480,exemplarsPerSeries=5,mmappedChunkT=3800-8 12.1s ± 1% 3.8s ±17% -68.35% (p=0.008 n=5+5)
LoadWAL/batches=100,seriesPerBatch=1000,samplesPerSeries=480,exemplarsPerSeries=24,mmappedChunkT=3800-8 12.4s ± 1% 4.0s ±18% -67.71% (p=0.008 n=5+5)
```
Benchmarked on Linux
```
goos: linux
goarch: amd64
pkg: github.com/prometheus/prometheus/tsdb
cpu: 11th Gen Intel(R) Core(TM) i7-1165G7 @ 2.80GHz
```
Signed-off-by: Jesus Vazquez <jesus.vazquez@grafana.com>
* Job queue
This PR reimplements chan chunkWriteJob with custom buffered queue that should use less memory, because it doesn't preallocate entire buffer for maximum queue size at once. Instead it allocates individual "segments" with smaller size.
As elements are added to the queue, they fill individual segments. When elements are removed from the queue (and segments), empty segments can be thrown away. This doesn't change memory usage of the queue when it's full, but should decrease its memory footprint when it's empty (queue will keep max 1 segment in such case).
Signed-off-by: Peter Štibraný <pstibrany@gmail.com>
* Modify test to work with low resolution timer.
Signed-off-by: Peter Štibraný <pstibrany@gmail.com>
* Improve comments.
Signed-off-by: Peter Štibraný <pstibrany@gmail.com>
* Avoid gaps in in-order data after restart with out-of-order enabled
Signed-off-by: Ganesh Vernekar <ganeshvern@gmail.com>
* Fix tests, do the temporary patch only if OOO is enabled
Signed-off-by: Ganesh Vernekar <ganeshvern@gmail.com>
* Avoid Peter's confusion
Signed-off-by: Ganesh Vernekar <ganeshvern@gmail.com>
* Use latest OutOfOrderTimeWindow
Signed-off-by: Ganesh Vernekar <ganeshvern@gmail.com>
This implementation is based on this design doc:
https://docs.google.com/document/d/1Kppm7qL9C-BJB1j6yb6-9ObG3AbdZnFUBYPNNWwDBYM/edit?usp=sharing
This commit adds support to accept out-of-order ("OOO") sample into the TSDB
up to a configurable time allowance. If OOO is enabled, overlapping querying
are automatically enabled.
Signed-off-by: Ganesh Vernekar <ganeshvern@gmail.com>
Co-authored-by: Jesus Vazquez <jesus.vazquez@grafana.com>
Co-authored-by: Ganesh Vernekar <ganeshvern@gmail.com>
Co-authored-by: Dieter Plaetinck <dieter@grafana.com>
* dont waste space on the chunkRefMap
* add time factor
* add comments
* better readability
* add instrumentation and more comments
* formatting
* uppercase comments
* Address review feedback. Renamed "free" to "shrink" everywhere, updated comments and threshold to 1000.
* double space
Signed-off-by: Mauro Stettler <mauro.stettler@gmail.com>
Co-authored-by: Peter Štibraný <pstibrany@gmail.com>
Co-authored-by: Mauro Stettler <mauro.stettler@gmail.com>
To avoid building up data in memory, commit and make a new appender
periodically.
The number `commitAfter = 10000` was chosen arbitrarily; testing with
10x more or less gives slightly worse results.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
When restarting Prometheus I sometimes see:
caller=db.go:832 level=error component=tsdb msg="compaction failed" err="compact head: persist head block: 2 errors: populate block: context canceled; context canceled"
And prometheus_tsdb_compactions_failed_total metric gets incremented.
This makes it more difficult to write alerts based on
prometheus_tsdb_compactions_failed_total metric since any restart can
trigger it.
Signed-off-by: Łukasz Mierzwa <l.mierzwa@gmail.com>
The code sleeps for a short time to allow goroutines to finish, however
it seems the duration can be reduced a lot, speeding up the reading
process.
I checked using some WAL data from production, and the queue is almost
always empty at the time we enter `waitForIdle()` so there is no danger
of spinning in the tight loop.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
Instead of creating a new hashing object every time, call `crc32.Checksum`
which computes the answer without allocations.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
This PR reimplements chan chunkWriteJob with custom buffered queue that should use less memory, because it doesn't preallocate entire buffer for maximum queue size at once. Instead it allocates individual "segments" with smaller size.
As elements are added to the queue, they fill individual segments. When elements are removed from the queue (and segments), empty segments can be thrown away. This doesn't change memory usage of the queue when it's full, but should decrease its memory footprint when it's empty (queue will keep max 1 segment in such case).
* dont waste space on the chunkRefMap
* add time factor
* add comments
* better readability
* add instrumentation and more comments
* formatting
* uppercase comments
* Address review feedback. Renamed "free" to "shrink" everywhere, updated comments and threshold to 1000.
* double space
Signed-off-by: Mauro Stettler <mauro.stettler@gmail.com>
Co-authored-by: Peter Štibraný <pstibrany@gmail.com>
* Rename walDir parameter to dir
Signed-off-by: Matej Gera <matejgera@gmail.com>
* Improve NewQueueManager comment
Signed-off-by: Matej Gera <matejgera@gmail.com>
This commit fixes a typo when reporting an error that the the symbols
table size has been exceeded.
Signed-off-by: Filip Petkovski <filip.petkovsky@gmail.com>
On macOS, the TestTombstoneCleanRetentionLimitsRace performs very
poorly. It takes more than a second to write out one block, and as it
writes 400 of them, we run into the 10-minute test timeout frequently.
While this doesn't fix the actual performance issue, breaking each
iteration into a subtest makes the test pass reliably (because each
iteration comfortably finishes in under a minute).
Related report: https://groups.google.com/g/prometheus-developers/c/jxQ6Ayg6VJ4/m/03H_DS9PDAAJ
Signed-off-by: Matthias Rampke <matthias@prometheus.io>
"Labels is a sorted set of labels. Order has to be guaranteed upon
instantiation." says the comment, so fix all the tests that break this
rule.
For `BenchmarkLabelValuesWithMatchers()` and
`BenchmarkHeadLabelValuesWithMatchers()` the amount of work done changes
significantly if you put the labels in order, because all series refs
get neatly partitioned by the `tens` label, so I renamed the labels
to maintain the previous behaviour.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
* refactor: move from io/ioutil to io and os packages
* use fs.DirEntry instead of os.FileInfo after os.ReadDir
Signed-off-by: MOREL Matthieu <matthieu.morel@cnp.fr>
* Add a test with variable samples rate append
This test overflows the chunk created in memseries, and the total amount
of samples in the (only) mmapped chunk is 29, instead of the 65565
appended ones.
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
* Cut new chunk when rate prediction was wrong
When appending samples at a slow rate, and then appending at a higher
rate, the prediction we made to cut a new chunk is no longer valid.
Sometimes this can even cause an overflow in the chunk, if more samples
than uint16 can hold are appended.
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
* Improve comment on 2*samplesPerChunk
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
* Assert that all chunks have less than 240 samples
Also, trigger new chunk at 240, not at more than 240
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
* tsdb/agent: Ignore duplicate exemplars
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Make each exemplar unique in TestCommit
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Re-Trigger CI for Windows and UI-related steps
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Change test comment to properly re-trigger pipeline
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* Defer Close() calls for test agent and segment reader
Signed-off-by: Paschalis Tsilias <paschalist0@gmail.com>
* tsdb/agent: port grafana/agent#676grafana/agent#676 fixed an issue where a loading a WAL with multiple
segments may result in ref ID collision. The starting ref ID for new
series should be set to the highest ref ID across all series records
from all WAL segments.
This fixes an issue where the starting ref ID was incorrectly set to the
highest ref ID found in the newest segment, which may not have any ref
IDs at all if no series records have been appended to it yet.
Signed-off-by: Robert Fratto <robertfratto@gmail.com>
* tsdb/agent: update terminology (s/ref ID/nextRef)
Signed-off-by: Robert Fratto <robertfratto@gmail.com>
This commit disables some unused workflows on our CI.
Also uses grafana/regexp instead of regexp which is blackisted.
Also updates head_test TestHeadReadWriterRepair increasing
ChunkWriteQueueSize to 1 so that the chunk disk mapper uses the async
queue. This seems to be default behaviour in upstream prometheus and
without this option our test fails.
* Chunks replay skips chunks with unknown encodings
We've changed the logic of loadMmappedChunks to skip chunks that have
unknown encodings. To do so we've modified IterateAllChunks to accept an
extra encoding argument in the callback function.
Also added unit tests in the head and chunk disk mapper.
* Also add an unit test for the old chunk diskmapper
* s/createUnsupportedChunk/writeUnsupportedChunk/g
* tsdb: avoid slice-to-interface allocation in EnsureOrder
This is pulling the `seriesRefSlice` out of the loop, so the compiler
doesn't allocate a new one on the heap every time.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
* tsdb: use pointer type in Pool for EnsureOrder
As noted by staticcheck, Pool prefers the objects in the pool to have
pointer type. This is a little more fiddly to code, but avoids
allocation of a wrapper object every time a slice is put into the pool.
Removed a comment that said fixing this has a performance penalty: not
borne out by benchmarks.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
