### Describe Your Changes
Introduce the `-search.maxDeleteSeries` flag that limits the number of
time series that can be deleted with a single
`/api/v1/admin/tsdb/delete_series` call.
Currently, any number can be deleted and if the number is big (millions)
then the operation may result in unaccounted CPU and memory usage spikes
which in some cases may result in OOM kill (see #7027). The flag limits
the number to 30k by default and the users may override it if needed at
the vmstorage start time.
---------
Signed-off-by: Artem Fetishev <rtm@victoriametrics.com>
Co-authored-by: Nikolay <nik@victoriametrics.com>
'any' type is supported starting from Go1.18. Let's consistently use it
instead of 'interface{}' type across the code base, since `any` is easier to read than 'interface{}'.
This reverts commit 6e395048d3.
Reason for revert: the previous logic was correct.
The purpose of `-search.maxSamplesPerQuery` command-line flag is to limit the amounts of CPU resources,
which could be taken by a single query - see https://docs.victoriametrics.com/#resource-usage-limits .
VictoriaMetrics processes samples in blocks during querying - it reads the block, then unpacks it,
then filters out samples outside the selected time range. This means that it _spends CPU time_
on reading and unpacking of _all the samples_ in every block on the requested time range,
even if only a single sample per each block matches the given time range.
The previous logic was effectively limiting CPU time a single query could take.
The new logic fails limiting CPU time a single query could take in some pathological cases
when only a small fraction of samples per each requested block fit the requested time range.
This allows performing multiplication DoS-attacks by querying very narrow time ranges over historical blocks,
which tend to be full. For example, if the `-search.maxSamplesPerQuery` equals to a billion,
and the query requests a single sample out of 8K samples per each block, this means that the query
may unpack a billion of such blocks without exceeding the limit, e.g. it may unpack and process 8K*1e9=8e12 samples.
This is not what the resource usage limits were created for originally - see https://docs.victoriametrics.com/#resource-usage-limits
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/5851
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/pull/6464
* lib/storage: add ability to use downsampling for the given series filter
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* docs: add information about downsampling filters
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* docs: fix MetricsQL filter
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* lib/storage/downsampling: treat missing downsampling filter as a bug
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* lib/storage/part_header: verify correctness of downsampling filters when opening partition
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* lib/storage/downsampling: save only appliable rules in part metadata
Filter and save only rules which are appliable to partition based on MinTimestamp of stored data.
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* lib/storage/downsampling: update log messages for final dedup
Properly specify a reason of re-running deduplication for partition.
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* lib/storage: consistently use MaxTimestamp to determine deduplication/downsampling rules
Using MinTimestamp leads to applying downsampling to parts which are only partially covered by downsampling rule.
For example, partition covers range [1000-2000]. At t=2100 and rule offset 500 data with t=2100-500 => 1600 must be downsampled. The range check against MinTimestamp evaluates to true even though partition contains range which must not be downsampled - [1600:2000].
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* Follow-up
- Apply the first matching downsampling period if multiple filters match the given time series.
This allows fine-tuning the downsampling config for the specific needs.
- Take into account downsampling filters during search queries.
- Reduce the difference between community and enterprise branches. This should simplify further maintenance of these branches.
- Properly parse series filters with colons inside them.
- Document the feature at docs/CHANGELOG.md.
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4960
---------
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
This commit returns back limits for these endpoints, which have been removed at 5d66ee88bd ,
since it has been appeared that missing limits result in high CPU usage, while the introduced concurrency limiter
results in failed lightweight requests to these endpoints because of timeout when heavyweight requests are executed.
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/5055
This case is possible after a new brsPool is allocated. The fix is to verify whether len(brsPool) >= len(brs.brs)
before trying to append a new item to brsPool and sharing its contents with brs.brs.
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/5733
This reduces the number of memory allocations at the cost of possible memory usage increase,
since now different metric name strings may hold references to the previous byte slice.
This is good tradeoff, since ProcessSearchQuery is called in vmselect, and vmselect isn't usually limited by memory.
This change has been extracted from https://github.com/VictoriaMetrics/VictoriaMetrics/pull/5527
* app/vmselect: limit the number of parallel workers by 32
The change should improve performance and memory usage during query processing
on machines with big number of CPU cores. The number of parallel workers for
query processing is controlled via `-search.maxWorkersPerQuery` command-line flag.
By default, the number of workers is limited by the number of available CPU cores,
but not more than 32. The limit can be increased via `-search.maxWorkersPerQuery`.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* wip
- The `-search.maxWorkersPerQuery` command-line flag doesn't limit resource usage,
so move it from the `resource usage limits` to `troubleshooting` chapter at docs/Single-server-VictoriaMetrics.md
- Make more clear the description for the `-search.maxWorkersPerQuery` command-line flag
- Add the description of `-search.maxWorkersPerQuery` to docs/Cluster-VictoriaMetrics.md
- Limit the maximum value, which can be passed to `-search.maxWorkersPerQuery`, to GOMAXPROCS,
because bigger values may worsen query performance and increase CPU usage
- Improve the the description of the change at docs/CHANGELOG.md. Mark it as FEATURE instead of BUGFIX,
since it is closer to a feature than to a bugfix.
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/5087
---------
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
using `runtime.Gosched` requires acquiring global lock to check if there are any other goroutines to perform tasks. with the latest versions of runtime it can pause running goroutines automatically without requiring to call `Gosched` directly.
Updates #3966
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
Call runtime.Gosched() only when there is a work to steal from other workers.
Simplify the timeseriesWorker() and unpackWroker() code a bit by inlining stealTimeseriesWork() and stealUnpackWork().
This should reduce CPU usage when processing queries on systems with big number of CPU cores.
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3966
Previously the selected time series were split evenly among available CPU cores
for further processing - e.g unpacking the data and applying the given rollup
function to the unpacked data.
Some time series could be processed slower than others.
This could result in uneven work distribution among available CPU cores,
e.g. some CPU cores could complete their work sooner than others.
This could slow down query execution.
The new algorithm allows stealing time series to process from other CPU cores
when all the local work is done. This should reduce the maximum time
needed for query execution (aka tail latency).
The new algorithm should also scale better on systems with many CPU cores,
since every CPU processes locally assigned time series without inter-CPU communications.
The inter-CPU communications are used only when all the local work is finished
and the pending work from other CPUs needs to be stealed.
Unpack time series with less than 400K samples in the currently running goroutine.
Previously a new goroutine was being started for unpacking the samples.
This was requiring additional memory allocations.
Usually the number of blocks returned per each time series during queries is around 4.
So it is a good idea to pre-allocate 4 block references per time series
in order to reduce the number of memory allocations.
Reduce inter-CPU communications when processing the query over big number of time series.
This should improve performance for queries over big number of time series
on systems with many CPU cores.
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2896
Based on b596ac3745
Thanks to @zqyzyq for the idea.
- Use binary search instead of linear scan when locating the run of smallest timestamps
in blocks with intersected time ranges. This should improve performance
when merging blocks with big number of samples
- Skip samples with duplicate timestamps. This should increase query performance
in cluster version of VictoriaMetrics with the enabled replication.
