mirror of
https://github.com/VictoriaMetrics/VictoriaMetrics.git
synced 2025-01-30 15:22:07 +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
185 lines
3.4 KiB
Go
185 lines
3.4 KiB
Go
package streamaggr
|
|
|
|
import (
|
|
"strings"
|
|
"sync"
|
|
"unsafe"
|
|
|
|
"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
|
|
"github.com/cespare/xxhash/v2"
|
|
)
|
|
|
|
const dedupAggrShardsCount = 128
|
|
|
|
type dedupAggr struct {
|
|
shards []dedupAggrShard
|
|
}
|
|
|
|
type dedupAggrShard struct {
|
|
dedupAggrShardNopad
|
|
|
|
// The padding prevents false sharing on widespread platforms with
|
|
// 128 mod (cache line size) = 0 .
|
|
_ [128 - unsafe.Sizeof(dedupAggrShardNopad{})%128]byte
|
|
}
|
|
|
|
type dedupAggrShardNopad struct {
|
|
mu sync.Mutex
|
|
m map[string]*dedupAggrSample
|
|
}
|
|
|
|
type dedupAggrSample struct {
|
|
value float64
|
|
}
|
|
|
|
func newDedupAggr() *dedupAggr {
|
|
shards := make([]dedupAggrShard, dedupAggrShardsCount)
|
|
return &dedupAggr{
|
|
shards: shards,
|
|
}
|
|
}
|
|
|
|
func (da *dedupAggr) sizeBytes() uint64 {
|
|
n := uint64(unsafe.Sizeof(*da))
|
|
for i := range da.shards {
|
|
n += da.shards[i].sizeBytes()
|
|
}
|
|
return n
|
|
}
|
|
|
|
func (da *dedupAggr) itemsCount() uint64 {
|
|
n := uint64(0)
|
|
for i := range da.shards {
|
|
n += da.shards[i].itemsCount()
|
|
}
|
|
return n
|
|
}
|
|
|
|
func (das *dedupAggrShard) sizeBytes() uint64 {
|
|
das.mu.Lock()
|
|
n := uint64(unsafe.Sizeof(*das))
|
|
for k, s := range das.m {
|
|
n += uint64(len(k)) + uint64(unsafe.Sizeof(k)+unsafe.Sizeof(s)+unsafe.Sizeof(*s))
|
|
}
|
|
das.mu.Unlock()
|
|
return n
|
|
}
|
|
|
|
func (das *dedupAggrShard) itemsCount() uint64 {
|
|
das.mu.Lock()
|
|
n := uint64(len(das.m))
|
|
das.mu.Unlock()
|
|
return n
|
|
}
|
|
|
|
func (da *dedupAggr) pushSamples(samples []pushSample) {
|
|
pss := getPerShardSamples()
|
|
shards := pss.shards
|
|
for _, sample := range samples {
|
|
h := xxhash.Sum64(bytesutil.ToUnsafeBytes(sample.key))
|
|
idx := h % uint64(len(shards))
|
|
shards[idx] = append(shards[idx], sample)
|
|
}
|
|
for i, shardSamples := range shards {
|
|
if len(shardSamples) == 0 {
|
|
continue
|
|
}
|
|
da.shards[i].pushSamples(shardSamples)
|
|
}
|
|
putPerShardSamples(pss)
|
|
}
|
|
|
|
type dedupFlushCtx struct {
|
|
samples []pushSample
|
|
}
|
|
|
|
func (ctx *dedupFlushCtx) reset() {
|
|
clear(ctx.samples)
|
|
ctx.samples = ctx.samples[:0]
|
|
}
|
|
|
|
func (da *dedupAggr) flush(f func(samples []pushSample)) {
|
|
ctx := &dedupFlushCtx{}
|
|
shards := da.shards
|
|
for i := range shards {
|
|
ctx.reset()
|
|
shards[i].flush(ctx, f)
|
|
}
|
|
}
|
|
|
|
type perShardSamples struct {
|
|
shards [][]pushSample
|
|
}
|
|
|
|
func (pss *perShardSamples) reset() {
|
|
shards := pss.shards
|
|
for i, shardSamples := range shards {
|
|
if len(shardSamples) > 0 {
|
|
clear(shardSamples)
|
|
shards[i] = shardSamples[:0]
|
|
}
|
|
}
|
|
}
|
|
|
|
func getPerShardSamples() *perShardSamples {
|
|
v := perShardSamplesPool.Get()
|
|
if v == nil {
|
|
return &perShardSamples{
|
|
shards: make([][]pushSample, dedupAggrShardsCount),
|
|
}
|
|
}
|
|
return v.(*perShardSamples)
|
|
}
|
|
|
|
func putPerShardSamples(pss *perShardSamples) {
|
|
pss.reset()
|
|
perShardSamplesPool.Put(pss)
|
|
}
|
|
|
|
var perShardSamplesPool sync.Pool
|
|
|
|
func (das *dedupAggrShard) pushSamples(samples []pushSample) {
|
|
das.mu.Lock()
|
|
defer das.mu.Unlock()
|
|
|
|
m := das.m
|
|
if m == nil {
|
|
m = make(map[string]*dedupAggrSample, len(samples))
|
|
das.m = m
|
|
}
|
|
for _, sample := range samples {
|
|
s, ok := m[sample.key]
|
|
if ok {
|
|
s.value = sample.value
|
|
} else {
|
|
key := strings.Clone(sample.key)
|
|
m[key] = &dedupAggrSample{
|
|
value: sample.value,
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
func (das *dedupAggrShard) flush(ctx *dedupFlushCtx, f func(samples []pushSample)) {
|
|
das.mu.Lock()
|
|
|
|
m := das.m
|
|
das.m = nil
|
|
|
|
das.mu.Unlock()
|
|
|
|
if len(m) == 0 {
|
|
return
|
|
}
|
|
|
|
dstSamples := ctx.samples
|
|
for key, s := range m {
|
|
dstSamples = append(dstSamples, pushSample{
|
|
key: key,
|
|
value: s.value,
|
|
})
|
|
}
|
|
ctx.samples = dstSamples
|
|
|
|
f(dstSamples)
|
|
}
|