lib/logstorage: use simpler in-memory cache instead of workingsetcache for caching recently ingested _stream values and recently queried set of streams

These caches aren't expected to grow big, so it is OK to use the most simplest cache based on sync.Map.
The benefit of this cache compared to workingsetcache is better scalability on systems with many CPU cores,
since it doesn't use mutexes at fast path.
An additional benefit is lower memory usage on average, since the size of in-memory cache equals
working set for the last 3 minutes.

The downside is that there is no upper bound for the cache size, so it may grow big during workload spikes.
But this is very unlikely for typical workloads.
This commit is contained in:
Aliaksandr Valialkin 2024-10-18 02:15:03 +02:00
parent 8aa144fa74
commit 0f24078146
No known key found for this signature in database
GPG key ID: 52C003EE2BCDB9EB
6 changed files with 159 additions and 26 deletions

83
lib/logstorage/cache.go Normal file
View file

@ -0,0 +1,83 @@
package logstorage
import (
"strings"
"sync"
"sync/atomic"
"time"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/bytesutil"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/timeutil"
)
type cache struct {
curr atomic.Pointer[sync.Map]
prev atomic.Pointer[sync.Map]
stopCh chan struct{}
wg sync.WaitGroup
}
func newCache() *cache {
var c cache
c.curr.Store(&sync.Map{})
c.prev.Store(&sync.Map{})
c.stopCh = make(chan struct{})
c.wg.Add(1)
go func() {
defer c.wg.Done()
c.runCleaner()
}()
return &c
}
func (c *cache) MustStop() {
close(c.stopCh)
c.wg.Wait()
}
func (c *cache) runCleaner() {
d := timeutil.AddJitterToDuration(3 * time.Minute)
t := time.NewTicker(d)
defer t.Stop()
for {
select {
case <-t.C:
c.clean()
case <-c.stopCh:
return
}
}
}
func (c *cache) clean() {
curr := c.curr.Load()
c.prev.Store(curr)
c.curr.Store(&sync.Map{})
}
func (c *cache) Get(k []byte) (any, bool) {
kStr := bytesutil.ToUnsafeString(k)
curr := c.curr.Load()
v, ok := curr.Load(kStr)
if ok {
return v, true
}
prev := c.prev.Load()
v, ok = prev.Load(kStr)
if ok {
kStr = strings.Clone(kStr)
curr.Store(kStr, v)
return v, true
}
return nil, false
}
func (c *cache) Set(k []byte, v any) {
kStr := string(k)
curr := c.curr.Load()
curr.Store(kStr, v)
}

View file

@ -0,0 +1,59 @@
package logstorage
import (
"fmt"
"testing"
)
func TestCache(t *testing.T) {
m := make(map[string]int)
for i := 0; i < 10; i++ {
k := fmt.Sprintf("key_%d", i)
m[k] = i
}
c := newCache()
defer c.MustStop()
for kStr := range m {
k := []byte(kStr)
if v, ok := c.Get(k); ok {
t.Fatalf("unexpected value obtained from the cache for key %q: %v", k, v)
}
c.Set(k, m[kStr])
v, ok := c.Get(k)
if !ok {
t.Fatalf("cannot obtain value for key %q", k)
}
if n := v.(int); n != m[kStr] {
t.Fatalf("unexpected value obtained for key %q; got %d; want %d", k, n, m[kStr])
}
}
// The cached entries should be still visible after a single clean() call.
c.clean()
for kStr := range m {
k := []byte(kStr)
v, ok := c.Get(k)
if !ok {
t.Fatalf("cannot obtain value for key %q", k)
}
if n := v.(int); n != m[kStr] {
t.Fatalf("unexpected value obtained for key %q; got %d; want %d", k, n, m[kStr])
}
}
// The cached entries must be dropped after two clean() calls.
c.clean()
c.clean()
for kStr := range m {
k := []byte(kStr)
if v, ok := c.Get(k); ok {
t.Fatalf("unexpected value obtained from the cache for key %q: %v", k, v)
}
}
}

View file

@ -500,13 +500,14 @@ func (idb *indexdb) marshalStreamFilterCacheKey(dst []byte, tenantIDs []TenantID
func (idb *indexdb) loadStreamIDsFromCache(tenantIDs []TenantID, sf *StreamFilter) ([]streamID, bool) {
bb := bbPool.Get()
bb.B = idb.marshalStreamFilterCacheKey(bb.B[:0], tenantIDs, sf)
data := idb.s.filterStreamCache.GetBig(nil, bb.B)
v, ok := idb.s.filterStreamCache.Get(bb.B)
bbPool.Put(bb)
if len(data) == 0 {
if !ok {
// Cache miss
return nil, false
}
// Cache hit - unpack streamIDs from data.
data := *(v.(*[]byte))
n, nSize := encoding.UnmarshalVarUint64(data)
if nSize <= 0 {
logger.Panicf("BUG: unexpected error when unmarshaling the number of streamIDs from cache")
@ -537,7 +538,7 @@ func (idb *indexdb) storeStreamIDsToCache(tenantIDs []TenantID, sf *StreamFilter
// Store marshaled streamIDs to cache.
bb := bbPool.Get()
bb.B = idb.marshalStreamFilterCacheKey(bb.B[:0], tenantIDs, sf)
idb.s.filterStreamCache.SetBig(bb.B, b)
idb.s.filterStreamCache.Set(bb.B, &b)
bbPool.Put(bb)
}

View file

@ -1,7 +1,6 @@
package logstorage
import (
"bytes"
"path/filepath"
"sort"
@ -160,20 +159,18 @@ func (pt *partition) logIngestedRows(lr *LogRows) {
}
func (pt *partition) hasStreamIDInCache(sid *streamID) bool {
var result [1]byte
bb := bbPool.Get()
bb.B = pt.marshalStreamIDCacheKey(bb.B, sid)
value := pt.s.streamIDCache.Get(result[:0], bb.B)
_, ok := pt.s.streamIDCache.Get(bb.B)
bbPool.Put(bb)
return bytes.Equal(value, okValue)
return ok
}
func (pt *partition) putStreamIDToCache(sid *streamID) {
bb := bbPool.Get()
bb.B = pt.marshalStreamIDCacheKey(bb.B, sid)
pt.s.streamIDCache.Set(bb.B, okValue)
pt.s.streamIDCache.Set(bb.B, nil)
bbPool.Put(bb)
}
@ -183,8 +180,6 @@ func (pt *partition) marshalStreamIDCacheKey(dst []byte, sid *streamID) []byte {
return dst
}
var okValue = []byte("1")
// debugFlush makes sure that all the recently ingested data data becomes searchable
func (pt *partition) debugFlush() {
pt.ddb.debugFlush()

View file

@ -6,7 +6,6 @@ import (
"time"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/timerpool"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/workingsetcache"
)
func TestPartitionLifecycle(t *testing.T) {
@ -183,8 +182,8 @@ func TestPartitionMustAddRowsConcurrent(t *testing.T) {
//
// When the storage is no longer needed, closeTestStorage() must be called.
func newTestStorage() *Storage {
streamIDCache := workingsetcache.New(1024 * 1024)
filterStreamCache := workingsetcache.New(1024 * 1024)
streamIDCache := newCache()
filterStreamCache := newCache()
return &Storage{
flushInterval: time.Second,
streamIDCache: streamIDCache,
@ -194,6 +193,6 @@ func newTestStorage() *Storage {
// closeTestStorage closes storage created via newTestStorage().
func closeTestStorage(s *Storage) {
s.streamIDCache.Stop()
s.filterStreamCache.Stop()
s.streamIDCache.MustStop()
s.filterStreamCache.MustStop()
}

View file

@ -11,9 +11,7 @@ import (
"github.com/VictoriaMetrics/VictoriaMetrics/lib/fs"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/memory"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/timeutil"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/workingsetcache"
)
// StorageStats represents stats for the storage. It may be obtained by calling Storage.UpdateStats().
@ -136,12 +134,12 @@ type Storage struct {
//
// It reduces the load on persistent storage during data ingestion by skipping
// the check whether the given stream is already registered in the persistent storage.
streamIDCache *workingsetcache.Cache
streamIDCache *cache
// filterStreamCache caches streamIDs keyed by (partition, []TenanID, StreamFilter).
//
// It reduces the load on persistent storage during querying by _stream:{...} filter.
filterStreamCache *workingsetcache.Cache
filterStreamCache *cache
}
type partitionWrapper struct {
@ -243,10 +241,8 @@ func MustOpenStorage(path string, cfg *StorageConfig) *Storage {
flockF := fs.MustCreateFlockFile(path)
// Load caches
mem := memory.Allowed()
streamIDCache := workingsetcache.New(mem / 16)
filterStreamCache := workingsetcache.New(mem / 10)
streamIDCache := newCache()
filterStreamCache := newCache()
s := &Storage{
path: path,
@ -463,10 +459,10 @@ func (s *Storage) MustClose() {
// between VictoriaLogs restarts. This may result in various issues
// during data ingestion and querying.
s.streamIDCache.Stop()
s.streamIDCache.MustStop()
s.streamIDCache = nil
s.filterStreamCache.Stop()
s.filterStreamCache.MustStop()
s.filterStreamCache = nil
// release lock file