app/vmselect:add command-line flag -search.inmemoryBufSizeBytes (#6869)

add command-line flag `-search.inmemoryBufSizeBytes` for configuring size of in-memory buffers used by vmselect during processing of vmstorage responses. A new summary metric `vm_tmp_blocks_inmemory_file_size_bytes` is exposed to show the size of the buffer during requests processing.

The new setting can be used by experienced users to adjust memory usage by vmselect when processing
many small read requests. Instead of allocating 4MB buffers each time, vmselect can be instructed to lower
the buffer size via `-search.inmemoryBufSizeBytes`. To make the decision whether this flag needs to be adjusted
users can consult with `vm_tmp_blocks_inmemory_file_size_bytes` which shows the actual size of buffers used
during query processing.

----------

The detailed information of this PR can be found in
https://github.com/VictoriaMetrics/VictoriaMetrics/pull/6851

### Checklist

The following checks are **mandatory**:

- [ ] My change adheres [VictoriaMetrics contributing
guidelines](https://docs.victoriametrics.com/contributing/).

---------

Co-authored-by: hagen1778 <roman@victoriametrics.com>
(cherry picked from commit cab3ef8294)

app/vmselect/netstorage/tmp_blocks_file.go

@@ -7,6 +7,7 @@ import (
 	"sync"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
+	"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/memory"
@@ -14,6 +15,10 @@ import (
 	"github.com/VictoriaMetrics/metrics"
 )
 
+var tmpBufSize = flagutil.NewBytes("search.inmemoryBufSizeBytes", 0, "Size for in-memory data blocks used during processing search requests. "+
+	"By default, the size is automatically calculated based on available memory. "+
+	"Adjust this flag value if you observe that vm_tmp_blocks_max_inmemory_file_size_bytes metric constantly shows much higher values than vm_tmp_blocks_inmemory_file_size_bytes. See https://github.com/VictoriaMetrics/VictoriaMetrics/pull/6851")
+
 // InitTmpBlocksDir initializes directory to store temporary search results.
 //
 // It stores data in system-defined temporary directory if tmpDirPath is empty.
@@ -29,6 +34,9 @@ func InitTmpBlocksDir(tmpDirPath string) {
 var tmpBlocksDir string
 
 func maxInmemoryTmpBlocksFile() int {
+	if tmpBufSize.IntN() > 0 {
+		return tmpBufSize.IntN()
+	}
 	mem := memory.Allowed()
 	maxLen := mem / 1024
 	if maxLen < 64*1024 {
@@ -40,9 +48,12 @@ func maxInmemoryTmpBlocksFile() int {
 	return maxLen
 }
 
-var _ = metrics.NewGauge(`vm_tmp_blocks_max_inmemory_file_size_bytes`, func() float64 {
+var (
-	return float64(maxInmemoryTmpBlocksFile())
+	_ = metrics.NewGauge(`vm_tmp_blocks_max_inmemory_file_size_bytes`, func() float64 {
-})
+		return float64(maxInmemoryTmpBlocksFile())
+	})
+	tmpBufSizeSummary = metrics.NewSummary(`vm_tmp_blocks_inmemory_file_size_bytes`)
+)
 
 type tmpBlocksFile struct {
 	buf []byte
@@ -65,6 +76,8 @@ func getTmpBlocksFile() *tmpBlocksFile {
 
 func putTmpBlocksFile(tbf *tmpBlocksFile) {
 	tbf.MustClose()
+	bufLen := tbf.Len()
+	tmpBufSizeSummary.Update(float64(bufLen))
 	tbf.buf = tbf.buf[:0]
 	tbf.f = nil
 	tbf.r = nil

docs/Cluster-VictoriaMetrics.md

@@ -1582,6 +1582,9 @@ Below is the output for `/path/to/vmselect -help`:
      Whether to fix lookback interval to 'step' query arg value. If set to true, the query model becomes closer to InfluxDB data model. If set to true, then -search.maxLookback and -search.maxStalenessInterval are ignored
   -search.skipSlowReplicas
      Whether to skip -replicationFactor - 1 slowest vmstorage nodes during querying. Enabling this setting may improve query speed, but it could also lead to incomplete results if some queried data has less than -replicationFactor copies at vmstorage nodes. Consider enabling this setting only if all the queried data contains -replicationFactor copies in the cluster
+  -search.inmemoryBufSizeBytes size
+     Size for in-memory data blocks used during processing search requests. By default, the size is automatically calculated based on available memory. Adjust this flag value if you observe that vm_tmp_blocks_max_inmemory_file_size_bytes metric constantly shows much higher values than vm_tmp_blocks_inmemory_file_size_bytes. See https://github.com/VictoriaMetrics/VictoriaMetrics/pull/6851
+     Supports the following optional suffixes for size values: KB, MB, GB, TB, KiB, MiB, GiB, TiB (default 0)
   -search.treatDotsAsIsInRegexps
      Whether to treat dots as is in regexp label filters used in queries. For example, foo{bar=~"a.b.c"} will be automatically converted to foo{bar=~"a\\.b\\.c"}, i.e. all the dots in regexp filters will be automatically escaped in order to match only dot char instead of matching any char. Dots in ".+", ".*" and ".{n}" regexps aren't escaped. This option is DEPRECATED in favor of {__graphite__="a.*.c"} syntax for selecting metrics matching the given Graphite metrics filter
   -selectNode array

docs/changelog/CHANGELOG.md

@@ -35,6 +35,7 @@ The value of `instance` label for those scrape targets will be changed from `<ad
 * FEATURE: [stream aggregation](https://docs.victoriametrics.com/stream-aggregation/): do not allow enabling `-stream.keepInput` and `keep_metric_names` options together in [stream aggregation config](https://docs.victoriametrics.com/stream-aggregation/#stream-aggregation-config), as it may result in time series collision.
 * FEATURE: [vmui](https://docs.victoriametrics.com/#vmui): add search functionality to the column display settings in the table. See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/6668).
 * FEATURE: [vmui](https://docs.victoriametrics.com/#vmui): add the ability to select all columns in the column display settings of the table. See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/6668). Thanks to @yincongcyincong for the [pull request](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/6680).
+* FEATURE: `vmselect` in [VictoriaMetrics cluster](https://docs.victoriametrics.com/cluster-victoriametrics/): add command-line flag `-search.inmemoryBufSizeBytes` for configuring size of in-memory buffers used by vmselect during processing of vmstorage responses. A new summary metric `vm_tmp_blocks_inmemory_file_size_bytes` is exposed to show the size of the buffer during requests processing. See this [pull request](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/6851) for details. Thanks to @tydhot for implementation.
 
 * BUGFIX: [vmagent](https://docs.victoriametrics.com/vmagent): fixes `proxy_url` authorization for scrape targets. Previously proxy authorization configuration was ignored for `https` targets. See [this](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/6771) issue for details.
 * BUGFIX: [vmagent](https://docs.victoriametrics.com/vmagent/) fix service discovery of Azure Virtual Machines for response contains `nextLink`. See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/6784).