This allows reducing the amounts of data, which must be read during queries over logs with big number of fields (aka "wide events").
This, in turn, improves query performance when the data, which needs to be scanned during the query, doesn't fit OS page cache.
This improves performance of `field_values` pipe when it is applied to large number of data blocks.
This also improves performance of /select/logsql/field_values HTTP API.
It is possible for in-memory part to be empty if ingested samples are
removed by retention filters. In this case, data will not be discarded
due to retention before creating in memory part. After in-memory parts
merge samples will be removed resulting in creating completely empty
part at destination.
This commit checks for resulting part and skips it, if it's empty.
---------
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
This commit adds Kubernetes Native Sidecar support.
It's the special type of init containers, that have restartPolicy == "Always" and continue to run after container initialization.
related issue https://github.com/VictoriaMetrics/VictoriaMetrics/issues/7287
Related issue:
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/7182
- add a separate index cache for searches which might read through large
amounts of random entries. Primary use-case for this is retention and
downsampling filters, when applying filters background merge needs to
fetch large amount of random entries which pollutes an index cache.
Using different caches allows to reduce effect on memory usage and cache
efficiency of the main cache while still having high cache hit rate. A
separate cache size is 5% of allowed memory.
- reduce size of indexdb/dataBlocks cache in order to free memory for
new sparse cache. Reduced size by 5% and moved this to a separate cache.
- add a separate metricName search which does not cache metric names -
this is needed in order to allow disabling metric name caching when
applying downsampling/retention filters. Applying filters during
background merge accesses random entries, this fills up cache and does
not provide an actual improvement due to random access nature.
Merge performance and memory usage stats before and after the change:
- before
- after
---------
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
This commit fixes the TestStorageRotateIndexDB flaky test reported at:
#6977. Sample test failure: https://pastebin.com/bTSs8HP1
The test fails because one goroutine adds items to the indexDB table
while another goroutine is closing that table. This may happen if
indexDB rotation happens twice during one Storage.add() operation:
- Storage.add() takes the current indexDB and adds index recods to it
- First index db rotation makes the current index DB a previous one
(still ok at this point)
- Second index db rotation removes the indexDB that was current two
rotations earlier. It does this by setting the mustDrop flag to true and
decrementing the ref counter. The ref counter reaches zero which cases
the underlying indexdb table to release its resources gracefully.
Graceful release assumes that the table is not written anymore. But
Storage.add() still adds items to it.
The solution is to increment the indexDB ref counters while it is used
inside add().
The unit test has been changed a little so that the test fails reliably.
The idea is to make add() function invocation to last much longer,
therefore the test inserts not just one record at a time but thouthands
of them.
To see the test fail, just replace the idbsLocked() func with:
```go
unc (s *Storage) idbsLocked2() (*indexDB, *indexDB, func()) {
return s.idbCurr.Load(), s.idbNext.Load(), func() {}
}
```
---------
Signed-off-by: Artem Fetishev <rtm@victoriametrics.com>
### Describe Your Changes
Add support for
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/6930
Calculate `-search.maxUniqueTimeseries` by
`-search.maxConcurrentRequests` and remaining memory if it's **not set**
or **less equal than 0**.
The remaining memory is affected by `-memory.allowedPercent`,
`-memory.allowedBytes` and cgroup memory limit.
### Checklist
The following checks are **mandatory**:
- [x] My change adheres [VictoriaMetrics contributing
guidelines](https://docs.victoriametrics.com/contributing/).
---------
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
(cherry picked from commit 85f60237e2)
Signed-off-by: hagen1778 <roman@victoriametrics.com>
These caches aren't expected to grow big, so it is OK to use the most simplest cache based on sync.Map.
The benefit of this cache compared to workingsetcache is better scalability on systems with many CPU cores,
since it doesn't use mutexes at fast path.
An additional benefit is lower memory usage on average, since the size of in-memory cache equals
working set for the last 3 minutes.
The downside is that there is no upper bound for the cache size, so it may grow big during workload spikes.
But this is very unlikely for typical workloads.
Partition directories can be manually deleted and copied from another sources such as backups or other VitoriaLogs instances.
In this case the persisted cache becomes out of sync with partitions. This can result in missing index entries
during data ingestion or in incorrect results during querying. So it is better to do not persist caches.
This shouldn't hurt VictoriaLogs performance just after the restart too much, since its caches usually contain
small amounts of data, which can be quickly re-populated from the persisted data.
Unpack the full columnsHeader block instead of unpacking meta-information per each individual column
when the query, which selects all the columns, is executed. This improves performance when scanning
logs with big number of fields.
- Use parallel merge of per-CPU shard results. This improves merge performance on multi-CPU systems.
- Use topN heap sort of per-shard results. This improves performance when results contain millions of entries.
The purpose of this change is to reduce confusion between using
`flag.Duration` and `flagutils.Duration`. The reason is that
`flagutils.Duration` was mistakenly used for cases that required `m`
support. See
ab0d31a7b0
The change in name should clearly indicate the purpose of this data
type.
### Describe Your Changes
Please provide a brief description of the changes you made. Be as
specific as possible to help others understand the purpose and impact of
your modifications.
### Checklist
The following checks are **mandatory**:
- [ ] My change adheres [VictoriaMetrics contributing
guidelines](https://docs.victoriametrics.com/contributing/).
Signed-off-by: hagen1778 <roman@victoriametrics.com>
### Describe Your Changes
`flagutil.Duration` docs state that `m` suffix stands for `minute`, but
in fact this suffix is not supported due to ambiguity with `month`
### Checklist
The following checks are **mandatory**:
- [x] My change adheres [VictoriaMetrics contributing
guidelines](https://docs.victoriametrics.com/contributing/).
Signed-off-by: Alexander Frolov <winningpiece@gmail.com>
### Describe Your Changes
If a dict flag has only one value without a prefix it is supposed to
replace default value.
Previously, when flag was set to `-flag=2` and the default value in
`NewDictInt` was set to 1 the resulting value for any `flag.Get()` call
would be 1 which is not expected.
This commit updates default value for the flag in case there is only one
entry for flag and the entry is a number without a key.
This affects cluster version and specifically `replicationFactor` flag
usage with vmstorage [node
groups](https://docs.victoriametrics.com/cluster-victoriametrics/#vmstorage-groups-at-vmselect).
Previously, the following configuration would effectively be ignored:
```
/path/to/vmselect \
-replicationFactor=2 \
-storageNode=g1/host1,g1/host2,g1/host3 \
-storageNode=g2/host4,g2/host5,g2/host6 \
-storageNode=g3/host7,g3/host8,g3/host9
```
Changes from this PR will force default value for `replicationFactor`
flag to be set to `2` which is expected as the result of this
configuration.
---------
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
1. Verify if field in [fields
pipe](https://docs.victoriametrics.com/victorialogs/logsql/#fields-pipe)
exists. If not, it generates a metric with illegal float value "" for
prometheus metrics protocol.
2. check if multiple time range filters produce conflicted query time
range, for instance:
```
query: _time: 5m | stats count(),
start:2024-10-08T10:00:00.806Z,
end: 2024-10-08T12:00:00.806Z,
time: 2024-10-10T10:02:59.806Z
```
must give no result due to invalid final time range.
---------
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
It has been appeared that VictoriaLogs is frequently used for collecting logs with tens of fields.
For example, standard Kuberntes setup on top of Filebeat generates more than 20 fields per each log.
Such logs are also known as "wide events".
The previous storage format was optimized for logs with a few fields. When at least a single field
was referenced in the query, then the all the meta-information about all the log fields was unpacked
and parsed per each scanned block during the query. This could require a lot of additional disk IO
and CPU time when logs contain many fields. Resolve this issue by providing an (field -> metainfo_offset)
index per each field in every data block. This index allows reading and extracting only the needed
metainfo for fields used in the query. This index is stored in columnsHeaderIndexFilename ( columns_header_index.bin ).
This allows increasing performance for queries over wide events by 10x and more.
Another issue was that the data for bloom filters and field values across all the log fields except of _msg
was intermixed in two files - fieldBloomFilename ( field_bloom.bin ) and fieldValuesFilename ( field_values.bin ).
This could result in huge disk read IO overhead when some small field was referred in the query,
since the Operating System usually reads more data than requested. It reads the data from disk
in at least 4KiB blocks (usually the block size is much bigger in the range 64KiB - 512KiB).
So, if 512-byte bloom filter or values' block is read from the file, then the Operating System
reads up to 512KiB of data from disk, which results in 1000x disk read IO overhead. This overhead isn't visible
for recently accessed data, since this data is usually stored in RAM (aka Operating System page cache),
but this overhead may become very annoying when performing the query over large volumes of data
which isn't present in OS page cache.
The solution for this issue is to split bloom filters and field values across multiple shards.
This reduces the worst-case disk read IO overhead by at least Nx where N is the number of shards,
while the disk read IO overhead is completely removed in best case when the number of columns doesn't exceed N.
Currently the number of shards is 8 - see bloomValuesShardsCount . This solution increases
performance for queries over large volumes of newly ingested data by up to 1000x.
The new storage format is versioned as v1, while the old storage format is version as v0.
It is stored in the partHeader.FormatVersion.
Parts with the old storage format are converted into parts with the new storage format during background merge.
It is possible to force merge by querying /internal/force_merge HTTP endpoint - see https://docs.victoriametrics.com/victorialogs/#forced-merge .
Empty fields are treated as non-existing fields by VictoriaLogs data model.
So there is no sense in returning empty fields in query results, since they may mislead and confuse users.
s.partitions can be changed when new partition is registered or when old partition is dropped.
This could lead to data races and panics when s.partitions slice is accessed by concurrently executed queries.
The fix is to make a copy of the selected partitions under s.partitionsLock before performing the query.
This localizes blockSearch.getColumnsHeader() call at block_search.go .
This call is going to be optimized in the next commits in order to avoid
unmarshaling of header data for unneeded columns, which weren't requested
by getConstColumnValue() / getColumnHeader().
Refer the original byte slice with the marshaled columnsHeader for columns names and dictionary-encoded column values.
This improves query performance a bit when big number of blocks with big number of columns are scanned during the query.
### Describe Your Changes
added opentelemetry exponential histograms support. Such histograms are automatically converted into
VictoriaMetrics histogram with `vmrange` buckets.
### Checklist
The following checks are **mandatory**:
- [ ] My change adheres [VictoriaMetrics contributing
guidelines](https://docs.victoriametrics.com/contributing/).
---------
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Co-authored-by: hagen1778 <roman@victoriametrics.com>
Improperly written pipes could be silently parsed as filter pipe.
For example, the following query:
* | by (x)
was silently parsed to:
* | filter "by" x
It is better to return error, so the user could identify and fix invalid pipe
instead of silently executing invalid query with `filter` pipe.
### Describe Your Changes
Currently, vmagent always uses a separate `http.Client` for every group
watcher in Kubernetes SD. With a high number of group watchers this
leads to large amount of opened connections.
This PR adds 2 changes to address this:
- re-use of existing `http.Client` - in case `http.Client` is connecting
to the same API server and uses the same parameters it will be re-used
between group watchers
- HTTP2 support - this allows to reuse connections more efficiently due
to ability of using streaming via existing connections.
See this issue for the details and test results -
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/5971
### Checklist
The following checks are **mandatory**:
- [ ] My change adheres [VictoriaMetrics contributing
guidelines](https://docs.victoriametrics.com/contributing/).
---------
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
This PR fixes#7062
For hijacked connections, one has to read from the connection buffer,
but still write directly to the connection. Otherwise, when reading
directly from such connections, the first byte may be lost. This, in
turn corrupts the ClientHello TLS handshake message and when the backend
server receives it, it closes the connection and reports the following
error in the log:
```
http: TLS handshake error from 127.0.0.1:33150: tls: first record does not look
like a TLS handshake
```
The first byte may be lost because underlying HTTP request handler may
read it from the connection and put it into the buffer. As the result,
subsequent connection reads won't see that byte.
- See: https://github.com/golang/go/issues/27408
- The fix is taken from : https://github.com/k3s-io/k3s/pull/6216
### Checklist
The following checks are **mandatory**:
- [x] My change adheres [VictoriaMetrics contributing
guidelines](https://docs.victoriametrics.com/contributing/).
Signed-off-by: Artem Fetishev <rtm@victoriametrics.com>
`vm_rows_ignored_total` metric is a metric for users to signalize about
ingestion issues, such as bad timestamp or parsing error.
In commit
a5424e95b3
this metric started to increment each time vmstorage gets NaN. But NaN
is a valid value for Prometheus data model and for Prometheus metrics
exposition format. Exporters from Prometheus ecosystem could expose NaNs
as values for metrics and these values will be delivered to vmstorage
and increment the metric.
Since there is nothing user can do with this, in opposite to parsing
errors or bad timestamps, there is not much sense in incrementing this
metric. So this commit rolls-back `reason="nan_value"` increments.
### Describe Your Changes
Please provide a brief description of the changes you made. Be as
specific as possible to help others understand the purpose and impact of
your modifications.
### Checklist
The following checks are **mandatory**:
- [ ] My change adheres [VictoriaMetrics contributing
guidelines](https://docs.victoriametrics.com/contributing/).
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Create blockResultColumn.forEachDictValue* helper functions for visiting matching
dictionary values. These helper functions should prevent from counting dictionary values
without matching logs in the future.
This is a follow-up for 0c0f013a60
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/7152
When ingesting samples with the same labels(duplicated samples or
samples with the same labels after `by` or `without` options). They
could register different entries for the same labelset in
LabelsCompressor.
For example, both index 99 and 100 can be assigned to label `foo=1` in
two concurrent pushes. Then due to differing label indexes in encoded
keys, the samples will appear as distinct in aggrState, resulting in
duplicated results after decompressing the label indexes.
fbde238cdc/lib/streamaggr/streamaggr.go (L933)
In this pull request, since we need to store `idxToLabel` first to
ensure the idx can be searched after `lc.labelToIdxStore`,
the `lc.idxToLabel` still could contain a duplicated entries
[100]="foo=1". But given the low likelihood of this issue and the size
of idxToLabel, it should be fine.
### 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>
Previously the phrase filter with `!` was treated unexpectedly.
For example, `foo!bar` filter was treated at `foo AND NOT bar`,
while most users expect that it matches "foo!bar" phrase.
This commit aligns with users' expectations.
