VictoriaMetrics/lib/streamaggr/avg.go
Aliaksandr Valialkin 28a9e92b5e
lib/streamaggr: huge pile of changes
- Reduce memory usage by up to 5x when de-duplicating samples across big number of time series.
- Reduce memory usage by up to 5x when aggregating across big number of output time series.
- Add lib/promutils.LabelsCompressor, which is going to be used by other VictoriaMetrics components
  for reducing memory usage for marshaled []prompbmarshal.Label.
- Add `dedup_interval` option at aggregation config, which allows setting individual
  deduplication intervals per each aggregation.
- Add `keep_metric_names` option at aggregation config, which allows keeping the original
  metric names in the output samples.
- Add `unique_samples` output, which counts the number of unique sample values.
- Add `increase_prometheus` and `total_prometheus` outputs, which ignore the first sample
  per each newly encountered time series.
- Use 64-bit hashes instead of marshaled labels as map keys when calculating `count_series` output.
  This makes obsolete https://github.com/VictoriaMetrics/VictoriaMetrics/pull/5579
- Expose various metrics, which may help debugging stream aggregation:
  - vm_streamaggr_dedup_state_size_bytes - the size of data structures responsible for deduplication
  - vm_streamaggr_dedup_state_items_count - the number of items in the deduplication data structures
  - vm_streamaggr_labels_compressor_size_bytes - the size of labels compressor data structures
  - vm_streamaggr_labels_compressor_items_count - the number of entries in the labels compressor
  - vm_streamaggr_flush_duration_seconds - a histogram, which shows the duration of stream aggregation flushes
  - vm_streamaggr_dedup_flush_duration_seconds - a histogram, which shows the duration of deduplication flushes
  - vm_streamaggr_flush_timeouts_total - counter for timed out stream aggregation flushes,
    which took longer than the configured interval
  - vm_streamaggr_dedup_flush_timeouts_total - counter for timed out deduplication flushes,
    which took longer than the configured dedup_interval
- Actualize docs/stream-aggregation.md

The memory usage reduction increases CPU usage during stream aggregation by up to 30%.

This commit is based on https://github.com/VictoriaMetrics/VictoriaMetrics/pull/5850
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/5898
2024-03-02 02:42:50 +02:00

81 lines
1.8 KiB
Go

package streamaggr
import (
"strings"
"sync"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
)
// avgAggrState calculates output=avg, e.g. the average value over input samples.
type avgAggrState struct {
m sync.Map
}
type avgStateValue struct {
mu sync.Mutex
sum float64
count int64
deleted bool
}
func newAvgAggrState() *avgAggrState {
return &avgAggrState{}
}
func (as *avgAggrState) 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 = &avgStateValue{
sum: s.value,
count: 1,
}
outputKey = strings.Clone(outputKey)
vNew, loaded := as.m.LoadOrStore(outputKey, v)
if !loaded {
// The entry has been successfully stored
continue
}
// Update the entry created by a concurrent goroutine.
v = vNew
}
sv := v.(*avgStateValue)
sv.mu.Lock()
deleted := sv.deleted
if !deleted {
sv.sum += s.value
sv.count++
}
sv.mu.Unlock()
if deleted {
// The entry has been deleted by the concurrent call to appendSeriesForFlush
// Try obtaining and updating the entry again.
goto again
}
}
}
func (as *avgAggrState) appendSeriesForFlush(ctx *flushCtx) {
currentTimeMsec := int64(fasttime.UnixTimestamp()) * 1000
m := &as.m
m.Range(func(k, v interface{}) bool {
// Atomically delete the entry from the map, so new entry is created for the next flush.
m.Delete(k)
sv := v.(*avgStateValue)
sv.mu.Lock()
avg := sv.sum / float64(sv.count)
// 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, "avg", currentTimeMsec, avg)
return true
})
}