Files
VictoriaMetrics/app/vminsert/common/streamaggr.go
Nikolay 11f488d8ff lib/streamaggr: concurrently push timeseries to aggregators
Previously all timeseries pushed into aggregators were added
sequentially. It could cause delays on data ingestion and it was not
possible to use all available.

 This commit adds concurrency based on available CPU cores.

Also, it adds new generic Buffer and BufferPool into slicesutil.

Fixes https://github.com/VictoriaMetrics/VictoriaMetrics/issues/9878
2025-10-24 10:29:48 +02:00

292 lines
10 KiB
Go

package common
import (
"flag"
"fmt"
"sync"
"sync/atomic"
"github.com/VictoriaMetrics/VictoriaMetrics/app/vmstorage"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/flagutil"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/procutil"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompb"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/slicesutil"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/storage"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/streamaggr"
"github.com/VictoriaMetrics/metrics"
)
var (
streamAggrConfig = flag.String("streamAggr.config", "", "Optional path to file with stream aggregation config. "+
"See https://docs.victoriametrics.com/victoriametrics/stream-aggregation/ . "+
"See also -streamAggr.keepInput, -streamAggr.dropInput and -streamAggr.dedupInterval")
streamAggrKeepInput = flag.Bool("streamAggr.keepInput", false, "Whether to keep all the input samples after the aggregation with -streamAggr.config. "+
"By default, only aggregated samples are dropped, while the remaining samples are stored in the database. "+
"See also -streamAggr.dropInput and https://docs.victoriametrics.com/victoriametrics/stream-aggregation/")
streamAggrDropInput = flag.Bool("streamAggr.dropInput", false, "Whether to drop all the input samples after the aggregation with -streamAggr.config. "+
"By default, only aggregated samples are dropped, while the remaining samples are stored in the database. "+
"See also -streamAggr.keepInput and https://docs.victoriametrics.com/victoriametrics/stream-aggregation/")
streamAggrDedupInterval = flag.Duration("streamAggr.dedupInterval", 0, "Input samples are de-duplicated with this interval before optional aggregation with -streamAggr.config . "+
"See also -streamAggr.dropInputLabels and -dedup.minScrapeInterval and https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#deduplication")
streamAggrDropInputLabels = flagutil.NewArrayString("streamAggr.dropInputLabels", "An optional list of labels to drop from samples "+
"before stream de-duplication and aggregation . See https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#dropping-unneeded-labels")
streamAggrIgnoreOldSamples = flag.Bool("streamAggr.ignoreOldSamples", false, "Whether to ignore input samples with old timestamps outside the current aggregation interval. "+
"See https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#ignoring-old-samples")
streamAggrIgnoreFirstIntervals = flag.Int("streamAggr.ignoreFirstIntervals", 0, "Number of aggregation intervals to skip after the start. Increase this value if you observe incorrect aggregation results after restarts. It could be caused by receiving unordered delayed data from clients pushing data into the database. "+
"See https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#ignore-aggregation-intervals-on-start")
streamAggrEnableWindows = flag.Bool("streamAggr.enableWindows", false, "Enables aggregation within fixed windows for all aggregators. "+
"This allows to get more precise results, but impacts resource usage as it requires twice more memory to store two states. "+
"See https://docs.victoriametrics.com/victoriametrics/stream-aggregation/#aggregation-windows.")
)
var (
saCfgReloaderStopCh chan struct{}
saCfgReloaderWG sync.WaitGroup
saCfgReloads *metrics.Counter
saCfgReloadErr *metrics.Counter
saCfgSuccess *metrics.Gauge
saCfgTimestamp *metrics.Counter
sasGlobal atomic.Pointer[streamaggr.Aggregators]
deduplicator *streamaggr.Deduplicator
)
// CheckStreamAggrConfig checks config pointed by -stramaggr.config
func CheckStreamAggrConfig() error {
if *streamAggrConfig == "" {
return nil
}
pushNoop := func(_ []prompb.TimeSeries) {}
opts := &streamaggr.Options{
DedupInterval: *streamAggrDedupInterval,
DropInputLabels: *streamAggrDropInputLabels,
IgnoreOldSamples: *streamAggrIgnoreOldSamples,
IgnoreFirstIntervals: *streamAggrIgnoreFirstIntervals,
EnableWindows: *streamAggrEnableWindows,
}
sas, err := streamaggr.LoadFromFile(*streamAggrConfig, pushNoop, opts, "global")
if err != nil {
return fmt.Errorf("error when loading -streamAggr.config=%q: %w", *streamAggrConfig, err)
}
sas.MustStop()
return nil
}
// InitStreamAggr must be called after flag.Parse and before using the common package.
//
// MustStopStreamAggr must be called when stream aggr is no longer needed.
func InitStreamAggr() {
saCfgReloaderStopCh = make(chan struct{})
if *streamAggrConfig == "" {
if *streamAggrDedupInterval > 0 {
deduplicator = streamaggr.NewDeduplicator(pushAggregateSeries, *streamAggrEnableWindows, *streamAggrDedupInterval, *streamAggrDropInputLabels, "global")
}
return
}
saCfgReloads = metrics.NewCounter(`vminsert_streamagg_config_reloads_total`)
saCfgReloadErr = metrics.NewCounter(`vminsert_streamagg_config_reloads_errors_total`)
saCfgSuccess = metrics.NewGauge(`vminsert_streamagg_config_last_reload_successful`, nil)
saCfgTimestamp = metrics.NewCounter(`vminsert_streamagg_config_last_reload_success_timestamp_seconds`)
sighupCh := procutil.NewSighupChan()
opts := &streamaggr.Options{
DedupInterval: *streamAggrDedupInterval,
DropInputLabels: *streamAggrDropInputLabels,
IgnoreOldSamples: *streamAggrIgnoreOldSamples,
IgnoreFirstIntervals: *streamAggrIgnoreFirstIntervals,
}
sas, err := streamaggr.LoadFromFile(*streamAggrConfig, pushAggregateSeries, opts, "global")
if err != nil {
logger.Fatalf("cannot load -streamAggr.config=%q: %s", *streamAggrConfig, err)
}
sasGlobal.Store(sas)
saCfgSuccess.Set(1)
saCfgTimestamp.Set(fasttime.UnixTimestamp())
// Start config reloader.
saCfgReloaderWG.Add(1)
go func() {
defer saCfgReloaderWG.Done()
for {
select {
case <-sighupCh:
case <-saCfgReloaderStopCh:
return
}
reloadStreamAggrConfig()
}
}()
}
func reloadStreamAggrConfig() {
logger.Infof("reloading -streamAggr.config=%q", *streamAggrConfig)
saCfgReloads.Inc()
opts := &streamaggr.Options{
DedupInterval: *streamAggrDedupInterval,
DropInputLabels: *streamAggrDropInputLabels,
IgnoreOldSamples: *streamAggrIgnoreOldSamples,
IgnoreFirstIntervals: *streamAggrIgnoreFirstIntervals,
}
sasNew, err := streamaggr.LoadFromFile(*streamAggrConfig, pushAggregateSeries, opts, "global")
if err != nil {
saCfgSuccess.Set(0)
saCfgReloadErr.Inc()
logger.Errorf("cannot reload -streamAggr.config=%q: use the previously loaded config; error: %s", *streamAggrConfig, err)
return
}
sas := sasGlobal.Load()
if !sasNew.Equal(sas) {
sasOld := sasGlobal.Swap(sasNew)
sasOld.MustStop()
logger.Infof("successfully reloaded stream aggregation config at -streamAggr.config=%q", *streamAggrConfig)
} else {
logger.Infof("nothing changed in -streamAggr.config=%q", *streamAggrConfig)
sasNew.MustStop()
}
saCfgSuccess.Set(1)
saCfgTimestamp.Set(fasttime.UnixTimestamp())
}
// MustStopStreamAggr stops stream aggregators.
func MustStopStreamAggr() {
close(saCfgReloaderStopCh)
saCfgReloaderWG.Wait()
sas := sasGlobal.Swap(nil)
sas.MustStop()
if deduplicator != nil {
deduplicator.MustStop()
deduplicator = nil
}
}
type streamAggrCtx struct {
mn storage.MetricName
tss []prompb.TimeSeries
labels []prompb.Label
samples []prompb.Sample
buf []byte
}
func (ctx *streamAggrCtx) Reset() {
ctx.mn.Reset()
clear(ctx.tss)
ctx.tss = ctx.tss[:0]
clear(ctx.labels)
ctx.labels = ctx.labels[:0]
ctx.samples = ctx.samples[:0]
ctx.buf = ctx.buf[:0]
}
func (ctx *streamAggrCtx) push(mrs []storage.MetricRow, matchIdxs []uint32) []uint32 {
mn := &ctx.mn
tss := ctx.tss
labels := ctx.labels
samples := ctx.samples
buf := ctx.buf
tssLen := len(tss)
for _, mr := range mrs {
if err := mn.UnmarshalRaw(mr.MetricNameRaw); err != nil {
logger.Panicf("BUG: cannot unmarshal recently marshaled MetricName: %s", err)
}
labelsLen := len(labels)
bufLen := len(buf)
buf = append(buf, mn.MetricGroup...)
metricGroup := bytesutil.ToUnsafeString(buf[bufLen:])
labels = append(labels, prompb.Label{
Name: "__name__",
Value: metricGroup,
})
for _, tag := range mn.Tags {
bufLen = len(buf)
buf = append(buf, tag.Key...)
name := bytesutil.ToUnsafeString(buf[bufLen:])
bufLen = len(buf)
buf = append(buf, tag.Value...)
value := bytesutil.ToUnsafeString(buf[bufLen:])
labels = append(labels, prompb.Label{
Name: name,
Value: value,
})
}
samplesLen := len(samples)
samples = append(samples, prompb.Sample{
Timestamp: mr.Timestamp,
Value: mr.Value,
})
tss = append(tss, prompb.TimeSeries{
Labels: labels[labelsLen:],
Samples: samples[samplesLen:],
})
}
ctx.tss = tss
ctx.labels = labels
ctx.samples = samples
ctx.buf = buf
tss = tss[tssLen:]
sas := sasGlobal.Load()
if sas.IsEnabled() {
matchIdxs = sas.Push(tss, matchIdxs)
} else if deduplicator != nil {
matchIdxs = slicesutil.SetLength(matchIdxs, len(tss))
for i := range matchIdxs {
matchIdxs[i] = 1
}
deduplicator.Push(tss)
}
ctx.Reset()
return matchIdxs
}
func pushAggregateSeries(tss []prompb.TimeSeries) {
currentTimestamp := int64(fasttime.UnixTimestamp()) * 1000
var ctx InsertCtx
ctx.Reset(len(tss))
ctx.skipStreamAggr = true
for _, ts := range tss {
labels := ts.Labels
ctx.Labels = ctx.Labels[:0]
for _, label := range labels {
name := label.Name
if name == "__name__" {
name = ""
}
ctx.AddLabel(name, label.Value)
}
value := ts.Samples[0].Value
if err := ctx.WriteDataPoint(nil, ctx.Labels, currentTimestamp, value); err != nil {
logger.Errorf("cannot store aggregate series: %s", err)
// Do not continue pushing the remaining samples, since it is likely they will return the same error.
return
}
}
// There is no need in limiting the number of concurrent calls to vmstorage.AddRows() here,
// since the number of concurrent pushAggregateSeries() calls should be already limited by lib/streamaggr.
if err := vmstorage.AddRows(ctx.mrs); err != nil {
logger.Errorf("cannot flush aggregate series: %s", err)
}
}