fix: Remove useless lock which cause porxy meta cache recursive lock (#30203)

relate: https://github.com/milvus-io/milvus/issues/30193

---------

Signed-off-by: aoiasd <zhicheng.yue@zilliz.com>
This commit is contained in:
aoiasd 2024-01-23 23:56:55 +08:00 committed by GitHub
parent cad8bf6c8d
commit 38746dfc2b
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 57 additions and 90 deletions

View File

@ -76,6 +76,7 @@ type Cache interface {
expireShardLeaderCache(ctx context.Context)
RemoveCollection(ctx context.Context, database, collectionName string)
RemoveCollectionsByID(ctx context.Context, collectionID UniqueID) []string
RemovePartition(ctx context.Context, database, collectionName string, partitionName string)
// GetCredentialInfo operate credential cache
GetCredentialInfo(ctx context.Context, username string) (*internalpb.CredentialInfo, error)
@ -102,8 +103,6 @@ type collectionInfo struct {
collID typeutil.UniqueID
schema *schemaInfo
partInfo *partitionInfos
leaderMutex sync.RWMutex
shardLeaders *shardLeaders
createdTimestamp uint64
createdUtcTimestamp uint64
consistencyLevel commonpb.ConsistencyLevel
@ -187,14 +186,6 @@ func (info *collectionInfo) isCollectionCached() bool {
return info != nil && info.collID != UniqueID(0) && info.schema != nil
}
func (info *collectionInfo) deprecateLeaderCache() {
info.leaderMutex.RLock()
defer info.leaderMutex.RUnlock()
if info.shardLeaders != nil {
info.shardLeaders.deprecated.Store(true)
}
}
// shardLeaders wraps shard leader mapping for iteration.
type shardLeaders struct {
idx *atomic.Int64
@ -248,13 +239,14 @@ type MetaCache struct {
rootCoord types.RootCoordClient
queryCoord types.QueryCoordClient
collInfo map[string]map[string]*collectionInfo // database -> collection -> collection_info
collInfo map[string]map[string]*collectionInfo // database -> collectionName -> collection_info
collLeader map[string]map[string]*shardLeaders // database -> collectionName -> collection_leaders
credMap map[string]*internalpb.CredentialInfo // cache for credential, lazy load
privilegeInfos map[string]struct{} // privileges cache
userToRoles map[string]map[string]struct{} // user to role cache
mu sync.RWMutex
credMut sync.RWMutex
privilegeMut sync.RWMutex
leaderMut sync.RWMutex
shardMgr shardClientMgr
sfGlobal conc.Singleflight[*collectionInfo]
}
@ -288,6 +280,7 @@ func NewMetaCache(rootCoord types.RootCoordClient, queryCoord types.QueryCoordCl
rootCoord: rootCoord,
queryCoord: queryCoord,
collInfo: map[string]map[string]*collectionInfo{},
collLeader: map[string]map[string]*shardLeaders{},
credMap: map[string]*internalpb.CredentialInfo{},
shardMgr: shardMgr,
privilegeInfos: map[string]struct{}{},
@ -318,6 +311,21 @@ func (m *MetaCache) getCollection(database, collectionName string, collectionID
return nil, false
}
func (m *MetaCache) getCollectionShardLeader(database, collectionName string) (*shardLeaders, bool) {
m.leaderMut.RLock()
defer m.leaderMut.RUnlock()
db, ok := m.collLeader[database]
if !ok {
return nil, false
}
if leaders, ok := db[collectionName]; ok {
return leaders, !leaders.deprecated.Load()
}
return nil, false
}
func (m *MetaCache) update(ctx context.Context, database, collectionName string, collectionID UniqueID) (*collectionInfo, error) {
if collInfo, ok := m.getCollection(database, collectionName, collectionID); ok {
return collInfo, nil
@ -355,20 +363,16 @@ func (m *MetaCache) update(ctx context.Context, database, collectionName string,
m.collInfo[database] = make(map[string]*collectionInfo)
}
_, ok := m.collInfo[database][collectionName]
if !ok {
m.collInfo[database][collectionName] = &collectionInfo{}
m.collInfo[database][collectionName] = &collectionInfo{
collID: collection.CollectionID,
schema: newSchemaInfo(collection.Schema),
partInfo: parsePartitionsInfo(infos),
createdTimestamp: collection.CreatedTimestamp,
createdUtcTimestamp: collection.CreatedUtcTimestamp,
consistencyLevel: collection.ConsistencyLevel,
}
collInfo := m.collInfo[database][collectionName]
collInfo.schema = newSchemaInfo(collection.Schema)
collInfo.collID = collection.CollectionID
collInfo.createdTimestamp = collection.CreatedTimestamp
collInfo.createdUtcTimestamp = collection.CreatedUtcTimestamp
collInfo.consistencyLevel = collection.ConsistencyLevel
collInfo.partInfo = parsePartitionsInfo(infos)
log.Info("meta update success", zap.String("database", database), zap.String("collectionName", collectionName), zap.Int64("collectionID", collInfo.collID))
log.Info("meta update success", zap.String("database", database), zap.String("collectionName", collectionName), zap.Int64("collectionID", collection.CollectionID))
return m.collInfo[database][collectionName], nil
}
@ -397,12 +401,10 @@ func (m *MetaCache) UpdateByID(ctx context.Context, database string, collectionI
// GetCollectionID returns the corresponding collection id for provided collection name
func (m *MetaCache) GetCollectionID(ctx context.Context, database, collectionName string) (UniqueID, error) {
method := "GetCollectionID"
m.mu.RLock()
collInfo, ok := m.getCollection(database, collectionName, 0)
if !ok {
metrics.ProxyCacheStatsCounter.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method, metrics.CacheMissLabel).Inc()
tr := timerecord.NewTimeRecorder("UpdateCache")
m.mu.RUnlock()
collInfo, err := m.UpdateByName(ctx, database, collectionName)
if err != nil {
@ -412,7 +414,6 @@ func (m *MetaCache) GetCollectionID(ctx context.Context, database, collectionNam
metrics.ProxyUpdateCacheLatency.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method).Observe(float64(tr.ElapseSpan().Milliseconds()))
return collInfo.collID, nil
}
defer m.mu.RUnlock()
metrics.ProxyCacheStatsCounter.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method, metrics.CacheHitLabel).Inc()
return collInfo.collID, nil
@ -421,13 +422,11 @@ func (m *MetaCache) GetCollectionID(ctx context.Context, database, collectionNam
// GetCollectionName returns the corresponding collection name for provided collection id
func (m *MetaCache) GetCollectionName(ctx context.Context, database string, collectionID int64) (string, error) {
method := "GetCollectionName"
m.mu.RLock()
collInfo, ok := m.getCollection(database, "", collectionID)
if !ok {
metrics.ProxyCacheStatsCounter.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method, metrics.CacheMissLabel).Inc()
tr := timerecord.NewTimeRecorder("UpdateCache")
m.mu.RUnlock()
collInfo, err := m.UpdateByID(ctx, database, collectionID)
if err != nil {
@ -437,21 +436,18 @@ func (m *MetaCache) GetCollectionName(ctx context.Context, database string, coll
metrics.ProxyUpdateCacheLatency.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method).Observe(float64(tr.ElapseSpan().Milliseconds()))
return collInfo.schema.Name, nil
}
defer m.mu.RUnlock()
metrics.ProxyCacheStatsCounter.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method, metrics.CacheHitLabel).Inc()
return collInfo.schema.Name, nil
}
func (m *MetaCache) GetCollectionInfo(ctx context.Context, database string, collectionName string, collectionID int64) (*collectionBasicInfo, error) {
m.mu.RLock()
collInfo, ok := m.getCollection(database, collectionName, 0)
method := "GetCollectionInfo"
// if collInfo.collID != collectionID, means that the cache is not trustable
// try to get collection according to collectionID
if !ok || collInfo.collID != collectionID {
m.mu.RUnlock()
tr := timerecord.NewTimeRecorder("UpdateCache")
metrics.ProxyCacheStatsCounter.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method, metrics.CacheMissLabel).Inc()
@ -462,7 +458,6 @@ func (m *MetaCache) GetCollectionInfo(ctx context.Context, database string, coll
metrics.ProxyUpdateCacheLatency.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method).Observe(float64(tr.ElapseSpan().Milliseconds()))
return collInfo.getBasicInfo(), nil
}
defer m.mu.RUnlock()
metrics.ProxyCacheStatsCounter.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method, metrics.CacheHitLabel).Inc()
return collInfo.getBasicInfo(), nil
@ -472,14 +467,12 @@ func (m *MetaCache) GetCollectionInfo(ctx context.Context, database string, coll
// If the information is not found, proxy will try to fetch information for other source (RootCoord for now)
// TODO: may cause data race of this implementation, should be refactored in future.
func (m *MetaCache) getFullCollectionInfo(ctx context.Context, database, collectionName string, collectionID int64) (*collectionInfo, error) {
m.mu.RLock()
collInfo, ok := m.getCollection(database, collectionName, collectionID)
method := "GetCollectionInfo"
// if collInfo.collID != collectionID, means that the cache is not trustable
// try to get collection according to collectionID
if !ok || collInfo.collID != collectionID {
m.mu.RUnlock()
tr := timerecord.NewTimeRecorder("UpdateCache")
metrics.ProxyCacheStatsCounter.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method, metrics.CacheMissLabel).Inc()
@ -491,18 +484,15 @@ func (m *MetaCache) getFullCollectionInfo(ctx context.Context, database, collect
return collInfo, nil
}
m.mu.RUnlock()
metrics.ProxyCacheStatsCounter.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method, metrics.CacheHitLabel).Inc()
return collInfo, nil
}
func (m *MetaCache) GetCollectionSchema(ctx context.Context, database, collectionName string) (*schemaInfo, error) {
m.mu.RLock()
collInfo, ok := m.getCollection(database, collectionName, 0)
method := "GetCollectionSchema"
if !ok {
m.mu.RUnlock()
tr := timerecord.NewTimeRecorder("UpdateCache")
metrics.ProxyCacheStatsCounter.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method, metrics.CacheMissLabel).Inc()
@ -516,7 +506,6 @@ func (m *MetaCache) GetCollectionSchema(ctx context.Context, database, collectio
zap.Int64("time (milliseconds) take ", tr.ElapseSpan().Milliseconds()))
return collInfo.schema, nil
}
defer m.mu.RUnlock()
metrics.ProxyCacheStatsCounter.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method, metrics.CacheHitLabel).Inc()
return collInfo.schema, nil
@ -566,12 +555,10 @@ func (m *MetaCache) GetPartitionsIndex(ctx context.Context, database, collection
}
func (m *MetaCache) GetPartitionInfos(ctx context.Context, database, collectionName string) (*partitionInfos, error) {
m.mu.RLock()
method := "GetPartitionInfo"
collInfo, ok := m.getCollection(database, collectionName, 0)
if !ok {
m.mu.RUnlock()
tr := timerecord.NewTimeRecorder("UpdateCache")
metrics.ProxyCacheStatsCounter.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method, metrics.CacheMissLabel).Inc()
@ -583,7 +570,6 @@ func (m *MetaCache) GetPartitionInfos(ctx context.Context, database, collectionN
metrics.ProxyUpdateCacheLatency.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method).Observe(float64(tr.ElapseSpan().Milliseconds()))
return collInfo.partInfo, nil
}
defer m.mu.RUnlock()
return collInfo.partInfo, nil
}
@ -796,6 +782,7 @@ func (m *MetaCache) UpdateCredential(credInfo *internalpb.CredentialInfo) {
// GetShards update cache if withCache == false
func (m *MetaCache) GetShards(ctx context.Context, withCache bool, database, collectionName string, collectionID int64) (map[string][]nodeInfo, error) {
method := "GetShards"
log := log.Ctx(ctx).With(
zap.String("collectionName", collectionName),
zap.Int64("collectionID", collectionID))
@ -805,16 +792,11 @@ func (m *MetaCache) GetShards(ctx context.Context, withCache bool, database, col
return nil, err
}
method := "GetShards"
cacheShardLeaders, ok := m.getCollectionShardLeader(database, collectionName)
if withCache {
var shardLeaders *shardLeaders
info.leaderMutex.RLock()
shardLeaders = info.shardLeaders
info.leaderMutex.RUnlock()
if shardLeaders != nil && !shardLeaders.deprecated.Load() {
if ok {
metrics.ProxyCacheStatsCounter.WithLabelValues(fmt.Sprint(paramtable.GetNodeID()), method, metrics.CacheHitLabel).Inc()
iterator := shardLeaders.GetReader()
iterator := cacheShardLeaders.GetReader()
return iterator.Shuffle(), nil
}
@ -830,36 +812,36 @@ func (m *MetaCache) GetShards(ctx context.Context, withCache bool, database, col
}
tr := timerecord.NewTimeRecorder("UpdateShardCache")
var resp *querypb.GetShardLeadersResponse
resp, err = m.queryCoord.GetShardLeaders(ctx, req)
resp, err := m.queryCoord.GetShardLeaders(ctx, req)
if err != nil {
return nil, err
}
if resp.GetStatus().GetErrorCode() != commonpb.ErrorCode_Success {
return nil, merr.Error(resp.GetStatus())
if err = merr.Error(resp.GetStatus()); err != nil {
return nil, err
}
shards := parseShardLeaderList2QueryNode(resp.GetShards())
info, err = m.getFullCollectionInfo(ctx, database, collectionName, collectionID)
if err != nil {
return nil, err
}
// lock leader
info.leaderMutex.Lock()
oldShards := info.shardLeaders
info.shardLeaders = &shardLeaders{
newShardLeaders := &shardLeaders{
shardLeaders: shards,
deprecated: atomic.NewBool(false),
idx: atomic.NewInt64(0),
}
iterator := info.shardLeaders.GetReader()
info.leaderMutex.Unlock()
// lock leader
m.leaderMut.Lock()
if _, ok := m.collLeader[database]; !ok {
m.collLeader[database] = make(map[string]*shardLeaders)
}
m.collLeader[database][collectionName] = newShardLeaders
m.leaderMut.Unlock()
iterator := newShardLeaders.GetReader()
ret := iterator.Shuffle()
oldLeaders := make(map[string][]nodeInfo)
if oldShards != nil {
oldLeaders = oldShards.shardLeaders
if cacheShardLeaders != nil {
oldLeaders = cacheShardLeaders.shardLeaders
}
// update refcnt in shardClientMgr
// and create new client for new leaders
@ -888,19 +870,8 @@ func parseShardLeaderList2QueryNode(shardsLeaders []*querypb.ShardLeadersList) m
// DeprecateShardCache clear the shard leader cache of a collection
func (m *MetaCache) DeprecateShardCache(database, collectionName string) {
log.Info("clearing shard cache for collection", zap.String("collectionName", collectionName))
m.mu.RLock()
var info *collectionInfo
var ok bool
db, dbOk := m.collInfo[database]
if !dbOk {
m.mu.RUnlock()
log.Warn("not found database", zap.String("dbName", database))
return
}
info, ok = db[collectionName]
m.mu.RUnlock()
if ok {
info.deprecateLeaderCache()
if shards, ok := m.getCollectionShardLeader(database, collectionName); ok {
shards.deprecated.Store(true)
}
}
@ -916,16 +887,16 @@ func (m *MetaCache) expireShardLeaderCache(ctx context.Context) {
log.Info("stop periodically update meta cache")
return
case <-ticker.C:
m.mu.RLock()
for database, db := range m.collInfo {
m.leaderMut.RLock()
for database, db := range m.collLeader {
log.RatedInfo(10, "expire all shard leader cache",
zap.String("database", database),
zap.Strings("collections", lo.Keys(db)))
for _, info := range db {
info.deprecateLeaderCache()
for _, shards := range db {
shards.deprecated.Store(true)
}
}
m.mu.RUnlock()
m.leaderMut.RUnlock()
}
}
}()

View File

@ -773,11 +773,6 @@ func TestMetaCache_ExpireShardLeaderCache(t *testing.T) {
err := InitMetaCache(ctx, rootCoord, queryCoord, shardMgr)
assert.NoError(t, err)
queryCoord.EXPECT().ShowCollections(mock.Anything, mock.Anything).Return(&querypb.ShowCollectionsResponse{
Status: merr.Success(),
CollectionIDs: []UniqueID{1},
InMemoryPercentages: []int64{100},
}, nil)
queryCoord.EXPECT().GetShardLeaders(mock.Anything, mock.Anything).Return(&querypb.GetShardLeadersResponse{
Status: merr.Success(),
Shards: []*querypb.ShardLeadersList{
@ -788,6 +783,7 @@ func TestMetaCache_ExpireShardLeaderCache(t *testing.T) {
},
},
}, nil)
nodeInfos, err := globalMetaCache.GetShards(ctx, true, dbName, "collection1", 1)
assert.NoError(t, err)
assert.Len(t, nodeInfos["channel-1"], 3)