mirror of
https://github.com/VictoriaMetrics/VictoriaMetrics.git
synced 2024-11-21 14:44:00 +00:00
0d5d46f9db
- 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
88 lines
2.2 KiB
Go
88 lines
2.2 KiB
Go
package streamaggr
|
|
|
|
import (
|
|
"strings"
|
|
"sync"
|
|
|
|
"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
|
|
"github.com/VictoriaMetrics/VictoriaMetrics/lib/fasttime"
|
|
"github.com/cespare/xxhash/v2"
|
|
)
|
|
|
|
// countSeriesAggrState calculates output=count_series, e.g. the number of unique series.
|
|
type countSeriesAggrState struct {
|
|
m sync.Map
|
|
}
|
|
|
|
type countSeriesStateValue struct {
|
|
mu sync.Mutex
|
|
m map[uint64]struct{}
|
|
deleted bool
|
|
}
|
|
|
|
func newCountSeriesAggrState() *countSeriesAggrState {
|
|
return &countSeriesAggrState{}
|
|
}
|
|
|
|
func (as *countSeriesAggrState) pushSamples(samples []pushSample) {
|
|
for i := range samples {
|
|
s := &samples[i]
|
|
inputKey, outputKey := getInputOutputKey(s.key)
|
|
|
|
// Count unique hashes over the inputKeys instead of unique inputKey values.
|
|
// This reduces memory usage at the cost of possible hash collisions for distinct inputKey values.
|
|
h := xxhash.Sum64(bytesutil.ToUnsafeBytes(inputKey))
|
|
|
|
again:
|
|
v, ok := as.m.Load(outputKey)
|
|
if !ok {
|
|
// The entry is missing in the map. Try creating it.
|
|
v = &countSeriesStateValue{
|
|
m: map[uint64]struct{}{
|
|
h: {},
|
|
},
|
|
}
|
|
outputKey = strings.Clone(outputKey)
|
|
vNew, loaded := as.m.LoadOrStore(outputKey, v)
|
|
if !loaded {
|
|
// The entry has been added to the map.
|
|
continue
|
|
}
|
|
// Update the entry created by a concurrent goroutine.
|
|
v = vNew
|
|
}
|
|
sv := v.(*countSeriesStateValue)
|
|
sv.mu.Lock()
|
|
deleted := sv.deleted
|
|
if !deleted {
|
|
if _, ok := sv.m[h]; !ok {
|
|
sv.m[h] = struct{}{}
|
|
}
|
|
}
|
|
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 *countSeriesAggrState) 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.(*countSeriesStateValue)
|
|
sv.mu.Lock()
|
|
n := len(sv.m)
|
|
// 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_series", currentTimeMsec, float64(n))
|
|
return true
|
|
})
|
|
}
|