Commit Graph

596 Commits

Author SHA1 Message Date
Dmytro Kozlov
f739e44802
lib/storage: fix returning camelcase label names (#3608)
* lib/storage: fix returning camelcase label names

* doc: add change log

* Update docs/CHANGELOG.md

* Update docs/CHANGELOG.md

Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
2023-01-07 01:11:10 -08:00
Aliaksandr Valialkin
b275983403
lib/writeconcurrencylimiter: improve the logic behind -maxConcurrentInserts limit
Previously the -maxConcurrentInserts was limiting the number of established client connections,
which write data to VictoriaMetrics. Some of these connections could be idle.
Such connections do not consume big amounts of CPU and RAM, so there is a little sense in limiting
the number of such connections. So now the -maxConcurrentInserts command-line option
limits the number of concurrently executed insert requests, not including idle connections.

It is recommended removing -maxConcurrentInserts command-line option, since the default value
for this option should work good for most cases.
2023-01-06 22:07:16 -08:00
Roman Khavronenko
57277ed6bc
vmstorage: add more context to the flock acquiring msg (#3584)
See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3578

Signed-off-by: hagen1778 <roman@victoriametrics.com>

Signed-off-by: hagen1778 <roman@victoriametrics.com>
2023-01-05 18:32:53 -08:00
Aliaksandr Valialkin
8dc04a86f6
lib/{storage,mergeset}: tune the threshold for assisted merge
The https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3425#issuecomment-1359117221
reveals that CPU usage for incoming queries may significantly increase when the number
of in-memory parts becomes too big.

This commit reduces the maximum number of in-memory parts before starting the assisted merge
during data ingestion. This should reduce CPU usage for incoming queries,
since they need to inspect lower number of in-memory parts.

This should help https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3425
2022-12-28 14:42:45 -08:00
Aliaksandr Valialkin
1ff62629f4
lib/storage: clear the err if it is set to io.EOF when searching for the TSID by metricID
This is expected error after when recently added indexdb data isn't available for search yet
or wasn't flushed to disk after unclean shutdown of VictoriaMetrics.

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3515
2022-12-20 14:05:53 -08:00
Aliaksandr Valialkin
2184de3bf2
lib/storage: do not check for the result returned by db.doExtDB() where this isn't necessary
This simplifies the code a bit
2022-12-19 13:23:30 -08:00
Aliaksandr Valialkin
11bd290201
lib/storage: search for TSIDs for the given metricIDs in the previous indexdb if they aren't found in the current indexdb
The issue triggers after the indexdb rotation for time series, which stop receiving new samples.
This results in missing data for such time series in query responses.

This commit should address the https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3502

The issue has been introduced in 2dd93449d8
2022-12-19 11:56:49 -08:00
Aliaksandr Valialkin
8c08d625ee
lib/storage: optimize partSearch.searchBHS() for common case when the TSID for the current block header is bigger or equal to the current tsid
This should help improving performance at https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3425
2022-12-19 10:31:39 -08:00
Aliaksandr Valialkin
512c73cef9
lib/storage: properly set buf capacity inside marshalMetricID
Previously it was always set to 0. In theory this could result into incorrect marshaling
of metricIDs.

The issue has been introduced in 5e4dfe50c6
2022-12-19 10:31:38 -08:00
Aliaksandr Valialkin
fbeebe4869
lib/storage: skip missing tsids in the current block header by using binary search
This improves performance by up to 10x when big number of the requested TSIDs
are missing in the searched parts.

This should help https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3425
2022-12-14 22:07:55 -08:00
Aliaksandr Valialkin
1a88fe5b1f
lib/flagutil/bytes.go: properly handle values bigger than 2GiB on 32-bit architectures
This fixes handling of values bigger than 2GiB for the following command-line flags:

- -storage.minFreeDiskSpaceBytes
- -remoteWrite.maxDiskUsagePerURL
2022-12-14 19:29:57 -08:00
Aliaksandr Valialkin
ea7940e5a7
lib/mergeset: reduce the parts threshold before starting assisted merges
This should improve query speed in general case.

This is a follow-up for d1af6046c7
2022-12-13 09:14:08 -08:00
Aliaksandr Valialkin
2a190f6451
lib/{mergeset,storage}: do not block small merges by pending big merges - assist with small merges instead
Blocked small merges may result into big number of small parts, which, in turn,
may result in increased CPU and memory usage during queries, since queries need to inspect
all the existing small parts.

The issue has been introduced in 8189770c50

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3337
2022-12-12 17:01:33 -08:00
Aliaksandr Valialkin
e56d5e1918
lib/storage: follow-up after 7c0ae3a86a
- Update docs at https://docs.victoriametrics.com/#deduplication
- Optimize the deduplication loop a bit

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3333
2022-12-08 18:18:49 -08:00
Roman Khavronenko
909cd04c55
lib/storage: keep sample with the biggest value on timestamp conflict (#3421)
The change leaves raw sample with the biggest value for identical
timestamps per each `-dedup.minScrapeInterval` discrete interval
when the deduplication is enabled.

```
benchstat old.txt new.txt
name                                         old time/op    new time/op    delta
DeduplicateSamples/minScrapeInterval=1s-10      817ns ± 2%     832ns ± 3%      ~     (p=0.052 n=10+10)
DeduplicateSamples/minScrapeInterval=2s-10     1.56µs ± 1%    2.12µs ± 0%   +35.19%  (p=0.000 n=9+7)
DeduplicateSamples/minScrapeInterval=5s-10     1.32µs ± 3%    1.65µs ± 2%   +25.57%  (p=0.000 n=10+10)
DeduplicateSamples/minScrapeInterval=10s-10    1.13µs ± 2%    1.50µs ± 1%   +32.85%  (p=0.000 n=10+10)

name                                         old speed      new speed      delta
DeduplicateSamples/minScrapeInterval=1s-10   10.0GB/s ± 2%   9.9GB/s ± 3%      ~     (p=0.052 n=10+10)
DeduplicateSamples/minScrapeInterval=2s-10   5.24GB/s ± 1%  3.87GB/s ± 0%   -26.03%  (p=0.000 n=9+7)
DeduplicateSamples/minScrapeInterval=5s-10   6.22GB/s ± 3%  4.96GB/s ± 2%   -20.37%  (p=0.000 n=10+10)
DeduplicateSamples/minScrapeInterval=10s-10  7.28GB/s ± 2%  5.48GB/s ± 1%   -24.74%  (p=0.000 n=10+10)
```

https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3333
Signed-off-by: hagen1778 <roman@victoriametrics.com>

Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-12-08 18:18:36 -08:00
Aliaksandr Valialkin
0a9992a9c6
lib/{storage,mergeset}: log the duration for flushing in-memory parts on graceful shutdown 2022-12-05 21:55:21 -08:00
Aliaksandr Valialkin
7d5c64eb7a
all: add -inmemoryDataFlushInterval command-line flag for controlling the frequency of saving in-memory data to disk
The main purpose of this command-line flag is to increase the lifetime of low-end flash storage
with the limited number of write operations it can perform. Such flash storage is usually
installed on Raspberry PI or similar appliances.

For example, `-inmemoryDataFlushInterval=1h` reduces the frequency of disk write operations
to up to once per hour if the ingested one-hour worth of data fits the limit for in-memory data.

The in-memory data is searchable in the same way as the data stored on disk.
VictoriaMetrics automatically flushes the in-memory data to disk on graceful shutdown via SIGINT signal.
The in-memory data is lost on unclean shutdown (hardware power loss, OOM crash, SIGKILL).

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3337
2022-12-05 15:28:09 -08:00
Aliaksandr Valialkin
9ac1174493
lib/{mergeset,storage}: add start background workers via startBackgroundWorkers() function 2022-12-04 00:01:14 -08:00
Aliaksandr Valialkin
a13d21513e
lib/mergeset: panic when too long item is passed to Table.AddItems() 2022-12-03 23:37:20 -08:00
Aliaksandr Valialkin
dccd70ce10
lib/storage: remove duplicate logging for filepath on errors 2022-12-03 23:15:28 -08:00
Aliaksandr Valialkin
813e8402f6
lib/storage: pass a single arg - rowsPerBlock - to getCompressLevel() function instead of two args 2022-12-03 23:10:26 -08:00
Aliaksandr Valialkin
bb93494eac
lib/{storage,mergeset}: use a single sync.WaitGroup for all background workers
This simplifies the code
2022-12-03 23:03:32 -08:00
Aliaksandr Valialkin
106332cd9f
lib/storage: properly pass retentionMsecs to OpenStorage() at TestIndexDBRepopulateAfterRotation 2022-12-03 23:03:30 -08:00
Aliaksandr Valialkin
ea55c16422
lib/{mergeset,storage}: pass compressLevel to blockStreamWriter.InitFromInmemoryPart
This allows packing in-memory blocks with different compression levels
depending on its contents. This may save memory usage.
2022-12-03 22:47:06 -08:00
Aliaksandr Valialkin
7ffa66d249
lib/{mergeset,storage}: take into account byte slice capacity when returning the size of in-memory part
This results in more correct reporting of memory usage for in-memory parts
2022-12-03 22:31:34 -08:00
Aliaksandr Valialkin
10a17bfa16
lib/{storage,mergeset}: consistency rename: `flushRaw{Rows,Items} -> flushPending{Rows,Items} 2022-12-03 22:18:05 -08:00
Aliaksandr Valialkin
233301a549
lib/storage: optimization: do not scan block for rows outside retention if it is covered by the retention 2022-12-03 22:14:20 -08:00
Aliaksandr Valialkin
fd9d0a550b
lib/storage: remove logging redundant path values in a single error message 2022-12-03 22:14:19 -08:00
Aliaksandr Valialkin
8e9822bc7f
lib/storage: speed up search for data block for the given tsids
Use binary search instead of linear scan for looking up the needed
data block inside index block.

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3425
2022-12-03 20:59:59 -08:00
Aliaksandr Valialkin
d8303845ef
lib/storage: fix TestUpdateCurrHourMetricIDs test when it runs on the first hour of the day by UTC 2022-12-02 17:23:59 -08:00
Aliaksandr Valialkin
d8d4d21d7a
lib/{mergeset,storage}: re-use the code for removing isInMerge flag at parts
Move the common code into releasePartsToMerge() method and consistently use it throughout the code.
2022-12-02 17:07:52 -08:00
匠心零度
d4808d5b84
lib/storage: remove extra error check (#3396) 2022-11-28 17:07:11 +01:00
Zakhar Bessarab
e407e7243a
{app/vmstorage,app/vmselect}: add API to get list of existing tenants (#3348)
* {app/vmstorage,app/vmselect}: add API to get list of existing tenants

* {app/vmstorage,app/vmselect}: add API to get list of existing tenants

* app/vmselect: fix error message

* {app/vmstorage,app/vmselect}: fix error messages

* app/vmselect: change log level for error handling

* wip

Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
2022-11-25 10:32:45 -08:00
Aliaksandr Valialkin
d3035b1ca1
lib/storage: follow-up for 790768f20b
- Document the bugfix at docs/CHANGELOG.md
- Simplify the bugfix a bit
2022-11-07 14:18:06 +02:00
Aliaksandr Valialkin
be78950011
lib/storage: typo fix after 32d48f8dfbb03174858c00bdfe6d9d22431dc8d8 2022-11-07 13:58:13 +02:00
Aliaksandr Valialkin
99e6a937a5
lib/storage: remove unused isFull field from hourMetricIDs struct 2022-11-07 13:15:59 +02:00
Aliaksandr Valialkin
ecb71a7221
lib/fs: add canOverwrite arg to WriteFileAtomically when it is allowed to overwrite the file atomically if it already exists 2022-10-26 01:08:35 +03:00
Aliaksandr Valialkin
a6d4711ac6
lib/storage: add support for retention filters (aka multiple retentions for distinct sets of time series)
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/143
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/289
2022-10-24 16:41:59 +03:00
Aliaksandr Valialkin
51f2e473f5
lib/storage: skip blocks outside the configured retention during search
Blocks outside the configured retention are eventually deleted during background merge.
But such blocks may reside in the storage for long time until background merge.
Previously VictoriaMetrics could spend additional CPU time on processing such blocks
during search queries. Now these blocks are skipped.
2022-10-24 02:56:13 +03:00
Aliaksandr Valialkin
2fc82b846e
lib/storage: do not pass retentionMsecs and isReadOnly args explicitly - access them via Storage arg
This makes code easier to read.

This is a follow-up after d2d30581a0
2022-10-24 01:32:56 +03:00
Aliaksandr Valialkin
d51f9b9284
lib/storage: small code cleanups 2022-10-24 01:17:58 +03:00
Aliaksandr Valialkin
5ace1587e6
lib/storage: re-use newTestStorage() instead of manually initializing Storage mock
This is a follow-up for d2d30581a0
2022-10-23 16:24:42 +03:00
Aliaksandr Valialkin
57ea7a3ee8
lib/storage: pass Storage to table and partition instead of getDeletedMetricIDs callback
This improves code readability a bit.
2022-10-23 16:11:02 +03:00
Aliaksandr Valialkin
63419d8e7c
lib/storage: small refactoring: move retentionDeadline to blockStreamMerger
This allows defining per-block retention in the future by updating the getRetentionDeadline function
2022-10-23 16:11:01 +03:00
Aliaksandr Valialkin
31071347ca
lib/storage: use a single reference to the currently merged block - bsm.Block during the block merge loop 2022-10-23 14:09:14 +03:00
Aliaksandr Valialkin
5d0a91afd5
lib/storage: properly pass uint64 constant to fmt.Errorf on 32-bit platforms 2022-10-23 12:48:43 +03:00
Aliaksandr Valialkin
2dd93449d8
lib/storage: subsitute searchTSIDs functions with more lightweight searchMetricIDs function
The searchTSIDs function was searching for metricIDs matching the the given tag filters
and then was locating the corresponding TSID entries for the found metricIDs.

The TSID entries aren't needed when searching for time series names (aka MetricName),
so this commit removes the uneeded TSID search from the implementation of /api/v1/series API.
This improves perfromance of /api/v1/series calls.

This commit also improves performance a bit for /api/v1/query and /api/v1/query_range calls,
since now these calls cache small metricIDs instead of big TSID entries
in the indexdb/tagFilters cache (now this cache is named indexdb/tagFiltersToMetricIDs)
without the need to compress the saved entries in order to save cache space.

This commit also removes concurrency limiter during searching for matching time series,
which was introduced in 8f16388428, since the concurrency
for all the read queries is already limited with -search.maxConcurrentRequests command-line flag.

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/648
2022-10-23 12:43:44 +03:00
Aliaksandr Valialkin
fe5611d6e1
lib/storage: free up memory occupied by Storage.pendingHourEntries after a temporary spike in its memory usage
This reduces vmstorage memory usage by up to 20% in production workload
2022-10-21 14:59:14 +03:00
Aliaksandr Valialkin
32b6ce691b
lib/storage: move common code to newRawRowsBlock() function 2022-10-21 14:46:06 +03:00
Aliaksandr Valialkin
2f8861ed9c
lib/storage: simplify code a bit after 3f5959c053 2022-10-21 14:39:44 +03:00