VictoriaMetrics/lib/streamaggr/count_samples.go
Aliaksandr Valialkin 32aa0751a1
lib/streamaggr: follow-up for 7cb894a777
- Use bytesutil.InternString() instead of strings.Clone() for inputKey and outputKey in aggregatorpushSamples().
  This should reduce string allocation rate, since strings can be re-used between aggrState flushes.
- Reduce memory allocations at dedupAggrShard by storing dedupAggrSample by value in the active series map.
- Remove duplicate call to bytesutil.InternBytes() at Deduplicator, since it is already called inside dedupAggr.pushSamples().
- Add missing string interning at rateAggrState.pushSamples().

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/pull/6402
2024-06-07 16:35:53 +02:00

83 lines
1.9 KiB
Go

package streamaggr
import (
"sync"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
)
// countSamplesAggrState calculates output=count_samples, e.g. the count of input samples.
type countSamplesAggrState struct {
m sync.Map
}
type countSamplesStateValue struct {
mu sync.Mutex
n uint64
deleted bool
}
func newCountSamplesAggrState() *countSamplesAggrState {
return &countSamplesAggrState{}
}
func (as *countSamplesAggrState) pushSamples(samples []pushSample) {
for i := range samples {
s := &samples[i]
outputKey := getOutputKey(s.key)
again:
v, ok := as.m.Load(outputKey)
if !ok {
// The entry is missing in the map. Try creating it.
v = &countSamplesStateValue{
n: 1,
}
outputKey = bytesutil.InternString(outputKey)
vNew, loaded := as.m.LoadOrStore(outputKey, v)
if !loaded {
// The new entry has been successfully created.
continue
}
// Use the entry created by a concurrent goroutine.
v = vNew
}
sv := v.(*countSamplesStateValue)
sv.mu.Lock()
deleted := sv.deleted
if !deleted {
sv.n++
}
sv.mu.Unlock()
if deleted {
// The entry has been deleted by the concurrent call to flushState
// Try obtaining and updating the entry again.
goto again
}
}
}
func (as *countSamplesAggrState) flushState(ctx *flushCtx, resetState bool) {
currentTimeMsec := int64(fasttime.UnixTimestamp()) * 1000
m := &as.m
m.Range(func(k, v interface{}) bool {
if resetState {
// Atomically delete the entry from the map, so new entry is created for the next flush.
m.Delete(k)
}
sv := v.(*countSamplesStateValue)
sv.mu.Lock()
n := sv.n
if resetState {
// Mark the entry as deleted, so it won't be updated anymore by concurrent pushSample() calls.
sv.deleted = true
}
sv.mu.Unlock()
key := k.(string)
ctx.appendSeries(key, "count_samples", currentTimeMsec, float64(n))
return true
})
}