milvus/internal/querycoordv2/handlers.go
chyezh a2502bde75
enhance: replica manager enhancement (#31496)
issue: #30647 

- ReplicaManager manage read only node now, and always do persistent of
node distribution of replica.

- All segment/channel checker using ReplicaManager to get read-only node
or read-write node, but not ResourceManager.

- ReplicaManager promise that only apply unique querynode to one replica
in same collection now (replicas in same collection never hold same
querynode at same time).

- ReplicaManager promise that fairly node count assignment policy if
multi replicas of collection is assigned to one resource group.

- Move some parameters check into ReplicaManager to avoid data race.

- Allow transfer replica to resource group that already load replica of
same collection

- Allow transfer node between resource groups that load replica of same
collection

---------

Signed-off-by: chyezh <chyezh@outlook.com>
2024-04-05 04:57:16 +08:00

455 lines
15 KiB
Go

// Licensed to the LF AI & Data foundation under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
// with the License. You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package querycoordv2
import (
"context"
"fmt"
"sync"
"time"
"github.com/cockroachdb/errors"
"github.com/samber/lo"
"go.uber.org/zap"
"github.com/milvus-io/milvus-proto/go-api/v2/commonpb"
"github.com/milvus-io/milvus-proto/go-api/v2/milvuspb"
"github.com/milvus-io/milvus/internal/proto/querypb"
"github.com/milvus-io/milvus/internal/querycoordv2/meta"
"github.com/milvus-io/milvus/internal/querycoordv2/session"
"github.com/milvus-io/milvus/internal/querycoordv2/task"
"github.com/milvus-io/milvus/internal/querycoordv2/utils"
"github.com/milvus-io/milvus/pkg/log"
"github.com/milvus-io/milvus/pkg/util/hardware"
"github.com/milvus-io/milvus/pkg/util/merr"
"github.com/milvus-io/milvus/pkg/util/metricsinfo"
"github.com/milvus-io/milvus/pkg/util/paramtable"
"github.com/milvus-io/milvus/pkg/util/typeutil"
"github.com/milvus-io/milvus/pkg/util/uniquegenerator"
)
// checkAnyReplicaAvailable checks if the collection has enough distinct available shards. These shards
// may come from different replica group. We only need these shards to form a replica that serves query
// requests.
func (s *Server) checkAnyReplicaAvailable(collectionID int64) bool {
for _, replica := range s.meta.ReplicaManager.GetByCollection(collectionID) {
isAvailable := true
for _, node := range replica.GetNodes() {
if s.nodeMgr.Get(node) == nil {
isAvailable = false
break
}
}
if isAvailable {
return true
}
}
return false
}
func (s *Server) getCollectionSegmentInfo(collection int64) []*querypb.SegmentInfo {
segments := s.dist.SegmentDistManager.GetByFilter(meta.WithCollectionID(collection))
currentTargetSegmentsMap := s.targetMgr.GetSealedSegmentsByCollection(collection, meta.CurrentTarget)
infos := make(map[int64]*querypb.SegmentInfo)
for _, segment := range segments {
if _, existCurrentTarget := currentTargetSegmentsMap[segment.GetID()]; !existCurrentTarget {
// if one segment exists in distMap but doesn't exist in currentTargetMap
// in order to guarantee that get segment request launched by sdk could get
// consistent result, for example
// sdk insert three segments:A, B, D, then A + B----compact--> C
// In this scenario, we promise that clients see either 2 segments(C,D) or 3 segments(A, B, D)
// rather than 4 segments(A, B, C, D), in which query nodes are loading C but have completed loading process
log.Info("filtered segment being in the intermediate status",
zap.Int64("segmentID", segment.GetID()))
continue
}
info, ok := infos[segment.GetID()]
if !ok {
info = &querypb.SegmentInfo{}
infos[segment.GetID()] = info
}
utils.MergeMetaSegmentIntoSegmentInfo(info, segment)
}
return lo.Values(infos)
}
// generate balance segment task and submit to scheduler
// if sync is true, this func call will wait task to finish, until reach the segment task timeout
// if copyMode is true, this func call will generate a load segment task, instead a balance segment task
func (s *Server) balanceSegments(ctx context.Context,
collectionID int64,
replica *meta.Replica,
srcNode int64,
dstNodes []int64,
segments []*meta.Segment,
sync bool,
copyMode bool,
) error {
log := log.Ctx(ctx).With(zap.Int64("collectionID", collectionID), zap.Int64("srcNode", srcNode))
plans := s.balancer.AssignSegment(collectionID, segments, dstNodes, true)
for i := range plans {
plans[i].From = srcNode
plans[i].Replica = replica
}
tasks := make([]task.Task, 0, len(plans))
for _, plan := range plans {
log.Info("manually balance segment...",
zap.Int64("replica", plan.Replica.GetID()),
zap.String("channel", plan.Segment.InsertChannel),
zap.Int64("from", plan.From),
zap.Int64("to", plan.To),
zap.Int64("segmentID", plan.Segment.GetID()),
)
actions := make([]task.Action, 0)
loadAction := task.NewSegmentActionWithScope(plan.To, task.ActionTypeGrow, plan.Segment.GetInsertChannel(), plan.Segment.GetID(), querypb.DataScope_Historical)
actions = append(actions, loadAction)
if !copyMode {
// if in copy mode, the release action will be skip
releaseAction := task.NewSegmentActionWithScope(plan.From, task.ActionTypeReduce, plan.Segment.GetInsertChannel(), plan.Segment.GetID(), querypb.DataScope_Historical)
actions = append(actions, releaseAction)
}
task, err := task.NewSegmentTask(s.ctx,
Params.QueryCoordCfg.SegmentTaskTimeout.GetAsDuration(time.Millisecond),
utils.ManualBalance,
collectionID,
plan.Replica,
actions...,
)
if err != nil {
log.Warn("create segment task for balance failed",
zap.Int64("replica", plan.Replica.GetID()),
zap.String("channel", plan.Segment.InsertChannel),
zap.Int64("from", plan.From),
zap.Int64("to", plan.To),
zap.Int64("segmentID", plan.Segment.GetID()),
zap.Error(err),
)
continue
}
task.SetReason("manual balance")
err = s.taskScheduler.Add(task)
if err != nil {
task.Cancel(err)
return err
}
tasks = append(tasks, task)
}
if sync {
err := task.Wait(ctx, Params.QueryCoordCfg.SegmentTaskTimeout.GetAsDuration(time.Millisecond), tasks...)
if err != nil {
msg := "failed to wait all balance task finished"
log.Warn(msg, zap.Error(err))
return errors.Wrap(err, msg)
}
}
return nil
}
// generate balance channel task and submit to scheduler
// if sync is true, this func call will wait task to finish, until reach the channel task timeout
// if copyMode is true, this func call will generate a load channel task, instead a balance channel task
func (s *Server) balanceChannels(ctx context.Context,
collectionID int64,
replica *meta.Replica,
srcNode int64,
dstNodes []int64,
channels []*meta.DmChannel,
sync bool,
copyMode bool,
) error {
log := log.Ctx(ctx).With(zap.Int64("collectionID", collectionID))
plans := s.balancer.AssignChannel(channels, dstNodes, true)
for i := range plans {
plans[i].From = srcNode
plans[i].Replica = replica
}
tasks := make([]task.Task, 0, len(plans))
for _, plan := range plans {
log.Info("manually balance channel...",
zap.Int64("replica", plan.Replica.GetID()),
zap.String("channel", plan.Channel.GetChannelName()),
zap.Int64("from", plan.From),
zap.Int64("to", plan.To),
)
actions := make([]task.Action, 0)
loadAction := task.NewChannelAction(plan.To, task.ActionTypeGrow, plan.Channel.GetChannelName())
actions = append(actions, loadAction)
if !copyMode {
// if in copy mode, the release action will be skip
releaseAction := task.NewChannelAction(plan.From, task.ActionTypeReduce, plan.Channel.GetChannelName())
actions = append(actions, releaseAction)
}
task, err := task.NewChannelTask(s.ctx,
Params.QueryCoordCfg.ChannelTaskTimeout.GetAsDuration(time.Millisecond),
utils.ManualBalance,
collectionID,
plan.Replica,
actions...,
)
if err != nil {
log.Warn("create channel task for balance failed",
zap.Int64("replica", plan.Replica.GetID()),
zap.String("channel", plan.Channel.GetChannelName()),
zap.Int64("from", plan.From),
zap.Int64("to", plan.To),
zap.Error(err),
)
continue
}
task.SetReason("manual balance")
err = s.taskScheduler.Add(task)
if err != nil {
task.Cancel(err)
return err
}
tasks = append(tasks, task)
}
if sync {
err := task.Wait(ctx, Params.QueryCoordCfg.ChannelTaskTimeout.GetAsDuration(time.Millisecond), tasks...)
if err != nil {
msg := "failed to wait all balance task finished"
log.Warn(msg, zap.Error(err))
return errors.Wrap(err, msg)
}
}
return nil
}
// TODO(dragondriver): add more detail metrics
func (s *Server) getSystemInfoMetrics(
ctx context.Context,
req *milvuspb.GetMetricsRequest,
) (string, error) {
clusterTopology := metricsinfo.QueryClusterTopology{
Self: metricsinfo.QueryCoordInfos{
BaseComponentInfos: metricsinfo.BaseComponentInfos{
Name: metricsinfo.ConstructComponentName(typeutil.QueryCoordRole, paramtable.GetNodeID()),
HardwareInfos: metricsinfo.HardwareMetrics{
IP: s.session.GetAddress(),
CPUCoreCount: hardware.GetCPUNum(),
CPUCoreUsage: hardware.GetCPUUsage(),
Memory: hardware.GetMemoryCount(),
MemoryUsage: hardware.GetUsedMemoryCount(),
Disk: hardware.GetDiskCount(),
DiskUsage: hardware.GetDiskUsage(),
},
SystemInfo: metricsinfo.DeployMetrics{},
CreatedTime: paramtable.GetCreateTime().String(),
UpdatedTime: paramtable.GetUpdateTime().String(),
Type: typeutil.QueryCoordRole,
ID: paramtable.GetNodeID(),
},
SystemConfigurations: metricsinfo.QueryCoordConfiguration{},
},
ConnectedNodes: make([]metricsinfo.QueryNodeInfos, 0),
}
metricsinfo.FillDeployMetricsWithEnv(&clusterTopology.Self.SystemInfo)
nodesMetrics := s.tryGetNodesMetrics(ctx, req, s.nodeMgr.GetAll()...)
s.fillMetricsWithNodes(&clusterTopology, nodesMetrics)
coordTopology := metricsinfo.QueryCoordTopology{
Cluster: clusterTopology,
Connections: metricsinfo.ConnTopology{
Name: metricsinfo.ConstructComponentName(typeutil.QueryCoordRole, paramtable.GetNodeID()),
// TODO(dragondriver): fill ConnectedComponents if necessary
ConnectedComponents: []metricsinfo.ConnectionInfo{},
},
}
resp, err := metricsinfo.MarshalTopology(coordTopology)
if err != nil {
return "", err
}
return resp, nil
}
func (s *Server) fillMetricsWithNodes(topo *metricsinfo.QueryClusterTopology, nodeMetrics []*metricResp) {
for _, metric := range nodeMetrics {
if metric.err != nil {
log.Warn("invalid metrics of query node was found",
zap.Error(metric.err))
topo.ConnectedNodes = append(topo.ConnectedNodes, metricsinfo.QueryNodeInfos{
BaseComponentInfos: metricsinfo.BaseComponentInfos{
HasError: true,
ErrorReason: metric.err.Error(),
// Name doesn't matter here because we can't get it when error occurs, using address as the Name?
Name: "",
ID: int64(uniquegenerator.GetUniqueIntGeneratorIns().GetInt()),
},
})
continue
}
if metric.resp.GetStatus().GetErrorCode() != commonpb.ErrorCode_Success {
log.Warn("invalid metrics of query node was found",
zap.Any("error_code", metric.resp.GetStatus().GetErrorCode()),
zap.Any("error_reason", metric.resp.GetStatus().GetReason()))
topo.ConnectedNodes = append(topo.ConnectedNodes, metricsinfo.QueryNodeInfos{
BaseComponentInfos: metricsinfo.BaseComponentInfos{
HasError: true,
ErrorReason: metric.resp.GetStatus().GetReason(),
Name: metric.resp.ComponentName,
ID: int64(uniquegenerator.GetUniqueIntGeneratorIns().GetInt()),
},
})
continue
}
infos := metricsinfo.QueryNodeInfos{}
err := metricsinfo.UnmarshalComponentInfos(metric.resp.Response, &infos)
if err != nil {
log.Warn("invalid metrics of query node was found",
zap.Error(err))
topo.ConnectedNodes = append(topo.ConnectedNodes, metricsinfo.QueryNodeInfos{
BaseComponentInfos: metricsinfo.BaseComponentInfos{
HasError: true,
ErrorReason: err.Error(),
Name: metric.resp.ComponentName,
ID: int64(uniquegenerator.GetUniqueIntGeneratorIns().GetInt()),
},
})
continue
}
topo.ConnectedNodes = append(topo.ConnectedNodes, infos)
}
}
type metricResp struct {
resp *milvuspb.GetMetricsResponse
err error
}
func (s *Server) tryGetNodesMetrics(ctx context.Context, req *milvuspb.GetMetricsRequest, nodes ...*session.NodeInfo) []*metricResp {
wg := sync.WaitGroup{}
ret := make([]*metricResp, 0, len(nodes))
retCh := make(chan *metricResp, len(nodes))
for _, node := range nodes {
node := node
wg.Add(1)
go func() {
defer wg.Done()
resp, err := s.cluster.GetMetrics(ctx, node.ID(), req)
if err != nil {
log.Warn("failed to get metric from QueryNode",
zap.Int64("nodeID", node.ID()))
return
}
retCh <- &metricResp{
resp: resp,
err: err,
}
}()
}
wg.Wait()
close(retCh)
for resp := range retCh {
ret = append(ret, resp)
}
return ret
}
func (s *Server) fillReplicaInfo(replica *meta.Replica, withShardNodes bool) (*milvuspb.ReplicaInfo, error) {
info := &milvuspb.ReplicaInfo{
ReplicaID: replica.GetID(),
CollectionID: replica.GetCollectionID(),
NodeIds: replica.GetNodes(),
ResourceGroupName: replica.GetResourceGroup(),
NumOutboundNode: s.meta.GetOutgoingNodeNumByReplica(replica),
}
channels := s.targetMgr.GetDmChannelsByCollection(replica.GetCollectionID(), meta.CurrentTarget)
if len(channels) == 0 {
msg := "failed to get channels, collection not loaded"
log.Warn(msg)
return nil, merr.WrapErrCollectionNotFound(replica.GetCollectionID(), msg)
}
var segments []*meta.Segment
if withShardNodes {
segments = s.dist.SegmentDistManager.GetByFilter(meta.WithCollectionID(replica.GetCollectionID()))
}
for _, channel := range channels {
leader, ok := s.dist.ChannelDistManager.GetShardLeader(replica, channel.GetChannelName())
var leaderInfo *session.NodeInfo
if ok {
leaderInfo = s.nodeMgr.Get(leader)
}
if leaderInfo == nil {
msg := fmt.Sprintf("failed to get shard leader for shard %s", channel)
log.Warn(msg)
return nil, merr.WrapErrNodeNotFound(leader, msg)
}
shard := &milvuspb.ShardReplica{
LeaderID: leader,
LeaderAddr: leaderInfo.Addr(),
DmChannelName: channel.GetChannelName(),
NodeIds: []int64{leader},
}
if withShardNodes {
shardNodes := lo.FilterMap(segments, func(segment *meta.Segment, _ int) (int64, bool) {
if replica.Contains(segment.Node) {
return segment.Node, true
}
return 0, false
})
shard.NodeIds = typeutil.NewUniqueSet(shardNodes...).Collect()
}
info.ShardReplicas = append(info.ShardReplicas, shard)
}
return info, nil
}
func filterDupLeaders(replicaManager *meta.ReplicaManager, leaders map[int64]*meta.LeaderView) map[int64]*meta.LeaderView {
type leaderID struct {
ReplicaID int64
Shard string
}
newLeaders := make(map[leaderID]*meta.LeaderView)
for _, view := range leaders {
replica := replicaManager.GetByCollectionAndNode(view.CollectionID, view.ID)
if replica == nil {
continue
}
id := leaderID{replica.GetID(), view.Channel}
if old, ok := newLeaders[id]; ok && old.Version > view.Version {
continue
}
newLeaders[id] = view
}
result := make(map[int64]*meta.LeaderView)
for _, v := range newLeaders {
result[v.ID] = v
}
return result
}