By holding a `proto.Buffer` per shard and passing it down to where
marshalling is done, we avoid creating a lot of garbage.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
* Moved everything to nPending buffer
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Simplify exemplar capacity addition
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Added pre-allocation
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Don't allocate if not sending exemplars
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Refactor: pass segment-reading function as param
To allow a different implementation to be used when garbage-collecting.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
* remote_write: reduce blocking from GC of series
Add a method `UpdateSeriesSegment()` which is used together with
`SeriesReset()` to garbage-collect old series. This allows us to
split the lock around queueManager series data and avoid blocking
`Append()` while reading series from the last checkpoint.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
* Cosmetic: review feedback on comments
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
* remote-write benchmark: include GC of series
Reduce the total number of samples per iteration from 5000*5000
(25 million) which is too big for my laptop, to 1*10000.
Extend `createTimeseries()` to add additional labels, so that the
queue manager is doing more realistic work.
Move the Append() call to a background goroutine - this works because
TestWriteClient uses a WaitGroup to signal completion.
Call `StoreSeries()` and `SeriesReset()` while adding samples, to
simulate the garbage-collection that wal.Watcher does.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
* Change BenchmarkSampleDelivery to call UpdateSeriesSegment
This matches what Watcher.garbageCollectSeries() is doing now
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
* Added MaxSamplesPerSend
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Added tests
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Fixed order of require
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Added docs
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* writes -> writesReceived
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Improved send loop
Signed-off-by: Levi Harrison <git@leviharrison.dev>
* Write exemplars to the WAL and send them over remote write.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Update example for exemplars, print data in a more obvious format.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Add metrics for remote write of exemplars.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Fix incorrect slices passed to send in remote write.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* We need to unregister the new metrics.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Address review comments
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Order of exemplar append vs write exemplar to WAL needs to change.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Several fixes to prevent sending uninitialized or incorrect samples with an exemplar. Fix dropping exemplar for missing series. Add tests for queue_manager sending exemplars
Signed-off-by: Martin Disibio <mdisibio@gmail.com>
* Store both samples and exemplars in the same timeseries buffer to remove the alloc when building final request, keep sub-slices in separate buffers for re-use
Signed-off-by: Martin Disibio <mdisibio@gmail.com>
* Condense sample/exemplar delivery tests to parameterized sub-tests
Signed-off-by: Martin Disibio <mdisibio@gmail.com>
* Rename test methods for clarity now that they also handle exemplars
Signed-off-by: Martin Disibio <mdisibio@gmail.com>
* Rename counter variable. Fix instances where metrics were not updated correctly
Signed-off-by: Martin Disibio <mdisibio@gmail.com>
* Add exemplars to LoadWAL benchmark
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* last exemplars timestamp metric needs to convert value to seconds with
ms precision
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Process exemplar records in a separate go routine when loading the WAL.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Address review comments related to clarifying comments and variable
names. Also refactor sample/exemplar to enqueue prompb types.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Regenerate types proto with comments, update protoc version again.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Put remote write of exemplars behind a feature flag.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Address some of Ganesh's review comments.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Move exemplar remote write feature flag to a config file field.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Address Bartek's review comments.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Don't allocate exemplar buffers in queue_manager if we're not going to
send exemplars over remote write.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Add ValidateExemplar function, validate exemplars when appending to head
and log them all to WAL before adding them to exemplar storage.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Address more reivew comments from Ganesh.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Add exemplar total label length check.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Address a few last review comments
Signed-off-by: Callum Styan <callumstyan@gmail.com>
Co-authored-by: Martin Disibio <mdisibio@gmail.com>
* Consider status code 429 as recoverable errors to avoid resharding.
* Adds support for Retry-After in backoff logic in remote storage.
Signed-off-by: Harkishen-Singh <harkishensingh@hotmail.com>
Currently there is no way of tracking the value of the
`max_samples_per_send` configuration option, which is commonly tweaked
when integrating with a remote write backend.
Signed-off-by: Jorge Luis Betancourt Gonzalez <jorge-luis.betancourt@trivago.com>
* Refactor global vars to avoid failure with run test more than once.
Signed-off-by: Harkishen-Singh <harkishensingh@hotmail.com>
* Register highestRecvTimestamp metric.
Signed-off-by: Harkishen-Singh <harkishensingh@hotmail.com>
* Use local interner vars.
Signed-off-by: Harkishen-Singh <harkishensingh@hotmail.com>
* Declare interner in write storage.
Signed-off-by: Harkishen-Singh <harkishensingh@hotmail.com>
* storage: Replace usage of sync/atomic with uber-go/atomic
Signed-off-by: Javier Palomo <javier.palomo.almena@gmail.com>
* tsdb: Replace usage of sync/atomic with uber-go/atomic
Signed-off-by: Javier Palomo <javier.palomo.almena@gmail.com>
* web: Replace usage of sync/atomic with uber-go/atomic
Signed-off-by: Javier Palomo <javier.palomo.almena@gmail.com>
* notifier: Replace usage of sync/atomic with uber-go/atomic
Signed-off-by: Javier Palomo <javier.palomo.almena@gmail.com>
* cmd: Replace usage of sync/atomic with uber-go/atomic
Signed-off-by: Javier Palomo <javier.palomo.almena@gmail.com>
* scripts: Verify that we are not using restricted packages
It checks that we are not directly importing 'sync/atomic'.
Signed-off-by: Javier Palomo <javier.palomo.almena@gmail.com>
* Reorganise imports in blocks
Signed-off-by: Javier Palomo <javier.palomo.almena@gmail.com>
* notifier/test: Apply PR suggestions
Signed-off-by: Javier Palomo <javier.palomo.almena@gmail.com>
* storage/remote: avoid storing references on newEntry
Signed-off-by: Javier Palomo <javier.palomo.almena@gmail.com>
* Revert "scripts: Verify that we are not using restricted packages"
This reverts commit 278d32748e.
Signed-off-by: Javier Palomo <javier.palomo.almena@gmail.com>
* web: Group imports accordingly
Signed-off-by: Javier Palomo <javier.palomo.almena@gmail.com>
Right now Queue Manager metrics are registered when the metrics struct
is created, which happens before a changed queue is shutdown and the old
metrics are unregistered. In the case of named queues or updates to
external labels the apply config will panic due to duplicate metrics.
Instead, register the metrics as part of starting the queue as we always
guarantee that Stop will be called before a new Start.
Signed-off-by: Chris Marchbanks <csmarchbanks@gmail.com>
* Pending Samples metric includes samples in channel
The pending samples metric should also include samples waiting in the
channels to be sent to provide a more accurate measure. In addition,
make sure that the pending samples is reset to 0 anytime a queue is
started as we remake all of the shards at that time.
Signed-off-by: Chris Marchbanks <csmarchbanks@gmail.com>
* Log the number of dropped samples on hard shutdown
Signed-off-by: Chris Marchbanks <csmarchbanks@gmail.com>
* Fixed nits introduced by https://github.com/prometheus/prometheus/pull/7334
* Added ChunkQueryable implementation to fanout and readyStorage.
* Added more comments.
* Changed NewVerticalChunkSeriesMerger to CompactingChunkSeriesMerger, removed tiny interface by reusing VerticalSeriesMergeFunc for overlapping algorithm for
both chunks and series, for both querying and compacting (!) + made sure duplicates are merged.
* Added ErrChunkSeriesSet
* Added Samples interface for seamless []promb.Sample to []tsdbutil.Sample conversion.
* Deprecating non chunks serieset based StreamChunkedReadResponses, added chunk one.
* Improved tests.
* Split remote client into Write (old storage) and read.
* Queryable client is now SampleAndChunkQueryable. Since we cannot use nice QueryableFunc I moved
all config based options to sampleAndChunkQueryableClient to aboid boilerplate.
In next commit: Changes for TSDB.
Signed-off-by: Bartlomiej Plotka <bwplotka@gmail.com>
* increase the remote write bucket range
Increase the range of remote write buckets to capture times above 10s for laggy scenarios
Buckets had been: {.005, .01, .025, .05, .1, .25, .5, 1, 2.5, 5, 10}
Buckets are now: {0.03125, 0.0625, 0.125, 0.25, 0.5, 1, 2, 4, 8, 16, 32, 64, 128, 256, 512}
Signed-off-by: Bert Hartmann <berthartm@gmail.com>
* revert back to DefBuckets with addons to be backwards compatible
Signed-off-by: Bert Hartmann <berthartm@gmail.com>
* shuffle the buckets to maintain 2-2.5x increases
Signed-off-by: Bert Hartmann <berthartm@gmail.com>
* Trace Remote Write requests
Signed-off-by: Cody Boggs <cboggs@splunk.com>
* Refactor store attempts to keep code flow clearer, and avoid so many places to deal with span finishing
Signed-off-by: Cody Boggs <cboggs@splunk.com>
Right now any new metrics added for remote write need to be added to
both the QueueManager struct, and the queueManagerMetrics struct.
Instead, use the queueManagerMetrics struct directly from QueueManager.
The newQueueManagerMetrics constructor will now create the metrics for a
specific queue with name and endpoint pre-populated, and a new copy of
the struct will be created specifically for each queue.
This also fixes a bug where prometheus_remote_storage_sent_bytes_total
is not being unregistered after a queue is changed.
Signed-off-by: Chris Marchbanks <csmarchbanks@gmail.com>
If the server is returning non-recoverable errors, such as if we are
trying to push samples that are too old, remote write will never
reshard. Non-recoverable errors should be treated the same as success
for the purpose of resharding, just as we do with sample rates and
durations.
Signed-off-by: Chris Marchbanks <csmarchbanks@gmail.com>
Remake the http client whenever ApplyConfig is called. This allows
secrets to be updated without needing to restart an otherwise unchanged
queue.
Signed-off-by: Chris Marchbanks <csmarchbanks@gmail.com>
* Fix bug with WAL watcher and Live Reader metrics usage.
Calling NewXMetrics when creating a Watcher or LiveReader results in a
registration error, which we're ignoring, and as a result other than the
first Watcher/Reader created, we had no metrics for either. So we would
only have metrics like Watcher Records Read for the first remote write
config in a users config file.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
The integral accumulator in the remote write sharding code is just a
second way of keeping track of the number of samples pending. Remove
integralAccumulator and use the samplesPending value we already
calculate to calculate the number of shards.
This has the added benefit of fixing a bug where the integralAccumulator
was not being initialized correctly due to not taking into account the
number of ticks being counted, causing the integralAccumulator initial
value to be off by an order of magnitude in some cases.
Signed-off-by: Chris Marchbanks <csmarchbanks@gmail.com>
* Track remote write queues via a map so we don't care about index.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Support a job name for remote write/read so we can differentiate between
them using the name.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Remote write/read has Name to not confuse the meaning of the field with
scrape job names.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Split queue/client label into remote_name and url labels.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Don't allow for duplicate remote write/read configs.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Ensure we restart remote write queues if the hash of their config has
not changed, but the remote name has changed.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Include name in remote read/write config hashes, simplify duplicates
check, update test accordingly.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
It is possible that desired shards is always a bit higher than the
number of shards (less than 30%) and by exporting desired shards as the
raw number it will be easy to tell if a Prometheus is in that situation.
Signed-off-by: Chris Marchbanks <csmarchbanks@gmail.com>
* Refactor calculateDesiredShards + don't reshard if we're having issues
sending samples.
* Track lastSendTimestamp via an int64 with atomic add/load, add a test
for reshard calculation.
* Simplify conditional for skipping resharding, add samplesIn/Out to shard
testcase struct.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
The WAL Watcher replays a checkpoint after it is created in order to
garbage collect series that no longer exist in the WAL. Currently the
garbage collection process is done serially with reading from the tip of
the WAL which can cause large delays in writing samples to remote
storage just after compaction occurs.
This also fixes a memory leak where dropped series are not cleaned up as
part of the SeriesReset process.
Signed-off-by: Chris Marchbanks <csmarchbanks@gmail.com>
The desired shards calculation now properly keeps track of the rate of
pending samples, and uses the previously unused integralAccumulator to
adjust for missing information in the desired shards calculation.
Also, configure more capacity for each shard. The default 10 capacity
causes shards to block on each other while
sending remote requests. Default to a 500 sample capacity and explain in
the documentation that having more capacity will help throughput.
Signed-off-by: Chris Marchbanks <csmarchbanks@gmail.com>
* Add benchmark for sample delivery
* Simplify StoreSeries to have only one loop
* Reduce allocations for pending samples in runShard
* Only allocate one send slice per segment
* Cache a buffer in each shard for snappy to use
* Remove queue manager seriesMtx
It is not possible for any of the places protected by the seriesMtx to
be called concurrently so it is safe to remove. By removing the mutex we
can simplify the Append code to one loop.
Signed-off-by: Chris Marchbanks <csmarchbanks@gmail.com>
* Don't panic if we try to release a string that is not in the interner.
* Move seriesMtx locking in QueueManager's StoreSeries function.
This stops us from calling release for strings that aren't interned if
there's a race between reading a checkpoint and storing new series
labels, which could happen during checkpointing or reloading config.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Unregister remote write queue manager specific metrics when stopping the
queue manager.
* Use DeleteLabelValues instead of Unregister to remove queue and watcher
related metrics when we stop them. Create those metrics in the structs
start functions rather than in their constructors because of the
ordering of creation, start, and stop in remote storage ApplyConfig.
* Add setMetrics function to WAL watcher so we can set
the watchers metrics in it's Start function, but not
have to call Start in some tests (causes data race).
Signed-off-by: Callum Styan <callumstyan@gmail.com>
a string when there are no longer any refs. Add tests for interning.
Co-authored-by: Tom Wilkie <tom.wilkie@gmail.com>
Signed-off-by: Callum Styan <callumstyan@gmail.com>
- Unmarshall external_labels config as labels.Labels, add tests.
- Convert some more uses of model.LabelSet to labels.Labels.
- Remove old relabel pkg (fixes#3647).
- Validate external label names.
Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>
* Consistently pre-lookup the metrics for a given queue in queue manager.
* Don't open the WAL (for writing) in the remote_write code.
* Add some more logging.
Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>
- Remove prometheus_remote_queue_last_send_timestamp_seconds metric. Its not particularly useful, we have highest_timestamp_seconds.
- Factor out maxGauage, a gauge that only increases.
- Change sharding calculations to use max samples in timestamp - max samples out timestamp (not rates).
- Also include the ratio of samples dropped to correctly predict number of pending samples.
Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>
- Add a dropped samples EWMA and use it in calculating desired shards.
- Update metric names and a log messages.
- Limit number of entries in the dedupe logging middleware to prevent potential OOM.
Signed-off-by: Callum Styan <callumstyan@gmail.com>
Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>
- If we're replaying the WAL to get series records, skip that segment when we hit corruptions.
- If we're tailing the WAL for samples, fail the watcher.
- When the watcher fails, restart from the latest checkpoint - and only send new samples by updating startTime.
- Tidy up log lines and error handling, don't return so many errors on quiting.
- Expect EOF when processing checkpoints.
Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>
- Use the queue name in WAL watcher logging.
- Don't return from watch if the reader error was EOF.
- Fix sample timestamp check logic regarding what samples we send.
- Refactor so we don't need readToEnd/readSeriesRecords
- Fix wal_watcher tests since readToEnd no longer exists
Signed-off-by: Callum Styan <callumstyan@gmail.com>
- Remove datarace in the exported highest scrape timestamp.
- Backoff on enqueue should be per-sample - reset the result for each sample.
- Remove diffKeys, unused ctx and cancelfunc in WALWatcher, 'name' from writeTo interface, and pass it to constructor.
- Reorder functions in WALWatcher depth-first according to call graph.
- Fix vendor/modules.txt.
- Split out the various timer periods into consts at the top of the file.
- Move w.currentSegmentMetric.Set close to where we set the currentSegment.
- Combine r.Next() and isClosed(w.quit) into a single loop.
- Unnest some ifs in WALWatcher.watch, propagate erros in decodeRecord, add some new lines to make it easier to read.
- Reorganise checkpoint handling to reduce nesting and make it easier to follow.
Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>
This change switches the remote_write API to use the TSDB WAL. This should reduce memory usage and prevent sample loss when the remote end point is down.
We use the new LiveReader from TSDB to tail WAL segments. Logic for finding the tracking segment is included in this PR. The WAL is tailed once for each remote_write endpoint specified. Reading from the segment is based on a ticker rather than relying on fsnotify write events, which were found to be complicated and unreliable in early prototypes.
Enqueuing a sample for sending via remote_write can now block, to provide back pressure. Queues are still required to acheive parallelism and batching. We have updated the queue config based on new defaults for queue capacity and pending samples values - much smaller values are now possible. The remote_write resharding code has been updated to prevent deadlocks, and extra tests have been added for these cases.
As part of this change, we attempt to guarantee that samples are not lost; however this initial version doesn't guarantee this across Prometheus restarts or non-retryable errors from the remote end (eg 400s).
This changes also includes the following optimisations:
- only marshal the proto request once, not once per retry
- maintain a single copy of the labels for given series to reduce GC pressure
Other minor tweaks:
- only reshard if we've also successfully sent recently
- add pending samples, latest sent timestamp, WAL events processed metrics
Co-authored-by: Chris Marchbanks <csmarchbanks.com> (initial prototype)
Co-authored-by: Tom Wilkie <tom.wilkie@gmail.com> (sharding changes)
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* *: use latest release of staticcheck
It also fixes a couple of things in the code flagged by the additional
checks.
Signed-off-by: Simon Pasquier <spasquie@redhat.com>
* Use official release of staticcheck
Also run 'go list' before staticcheck to avoid failures when downloading packages.
Signed-off-by: Simon Pasquier <spasquie@redhat.com>
More than one remote_write destination can be configured, in which
case it's essential to know which one each log message refers to.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
This can happen in the situation where the system scales up the number of shards massively (to deal with some backlog), then scales it down again as the number of samples sent during the time period is less than the number received.
sniper