wei liu 54619eaa2c
feat: Implement partial result support on node down (#42009)
issue: https://github.com/milvus-io/milvus/issues/41690
This commit implements partial search result functionality when query
nodes go down, improving system availability during node failures. The
changes include:

- Enhanced load balancing in proxy (lb_policy.go) to handle node
failures with retry support
- Added partial search result capability in querynode delegator and
distribution logic
- Implemented tests for various partial result scenarios when nodes go
down
- Added metrics to track partial search results in querynode_metrics.go
- Updated parameter configuration to support partial result required
data ratio
- Replaced old partial_search_test.go with more comprehensive
partial_result_on_node_down_test.go
- Updated proto definitions and improved retry logic

These changes improve query resilience by returning partial results to
users when some query nodes are unavailable, ensuring that queries don't
completely fail when a portion of data remains accessible.

---------

Signed-off-by: Wei Liu <wei.liu@zilliz.com>
2025-05-28 00:12:28 +08:00

252 lines
9.0 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 utils
import (
"context"
"fmt"
"time"
"go.uber.org/zap"
"github.com/milvus-io/milvus/internal/querycoordv2/meta"
"github.com/milvus-io/milvus/internal/querycoordv2/session"
"github.com/milvus-io/milvus/pkg/v2/log"
"github.com/milvus-io/milvus/pkg/v2/proto/querypb"
"github.com/milvus-io/milvus/pkg/v2/util/merr"
"github.com/milvus-io/milvus/pkg/v2/util/paramtable"
)
func CheckNodeAvailable(nodeID int64, info *session.NodeInfo) error {
if info == nil {
return merr.WrapErrNodeOffline(nodeID)
}
return nil
}
// In a replica, a shard is available, if and only if:
// 1. The leader is online
// 2. All QueryNodes in the distribution are online
// 3. The last heartbeat response time is within HeartbeatAvailableInterval for all QueryNodes(include leader) in the distribution
// 4. All segments of the shard in target should be in the distribution
func CheckDelegatorDataReady(nodeMgr *session.NodeManager, targetMgr meta.TargetManagerInterface, leader *meta.LeaderView, scope int32) error {
log := log.Ctx(context.TODO()).
WithRateGroup(fmt.Sprintf("util.CheckDelegatorDataReady-%d", leader.CollectionID), 1, 60).
With(zap.Int64("leaderID", leader.ID), zap.Int64("collectionID", leader.CollectionID))
// Check whether leader is online
info := nodeMgr.Get(leader.ID)
if info == nil {
err := merr.WrapErrNodeOffline(leader.ID)
log.Info("leader is not available", zap.Error(err))
return fmt.Errorf("leader not available: %w", err)
}
segmentDist := targetMgr.GetSealedSegmentsByChannel(context.TODO(), leader.CollectionID, leader.Channel, scope)
// Check whether segments are fully loaded
for segmentID := range segmentDist {
version, exist := leader.Segments[segmentID]
if !exist {
log.RatedInfo(10, "leader is not available due to lack of segment", zap.Int64("segmentID", segmentID))
return merr.WrapErrSegmentLack(segmentID)
}
// Check whether segment's worker node is online
info := nodeMgr.Get(version.GetNodeID())
if info == nil {
err := merr.WrapErrNodeOffline(leader.ID)
log.Info("leader is not available due to QueryNode unavailable",
zap.Int64("segmentID", segmentID),
zap.Error(err))
return err
}
}
return nil
}
func checkLoadStatus(ctx context.Context, m *meta.Meta, collectionID int64) error {
percentage := m.CollectionManager.CalculateLoadPercentage(ctx, collectionID)
if percentage < 0 {
err := merr.WrapErrCollectionNotLoaded(collectionID)
log.Ctx(ctx).Warn("failed to GetShardLeaders", zap.Error(err))
return err
}
collection := m.CollectionManager.GetCollection(ctx, collectionID)
if collection != nil && collection.GetStatus() == querypb.LoadStatus_Loaded {
// when collection is loaded, regard collection as readable, set percentage == 100
percentage = 100
}
if percentage < 100 {
err := merr.WrapErrCollectionNotFullyLoaded(collectionID)
msg := fmt.Sprintf("collection %v is not fully loaded", collectionID)
log.Ctx(ctx).Warn(msg)
return err
}
return nil
}
func GetShardLeadersWithChannels(
ctx context.Context,
m *meta.Meta,
dist *meta.DistributionManager,
nodeMgr *session.NodeManager,
collectionID int64,
channels map[string]*meta.DmChannel,
withUnserviceableShards bool,
) ([]*querypb.ShardLeadersList, error) {
ret := make([]*querypb.ShardLeadersList, 0)
replicas := m.ReplicaManager.GetByCollection(ctx, collectionID)
for _, channel := range channels {
log := log.Ctx(ctx).With(zap.String("channel", channel.GetChannelName()))
ids := make([]int64, 0, len(replicas))
addrs := make([]string, 0, len(replicas))
serviceable := make([]bool, 0, len(replicas))
for _, replica := range replicas {
leader := dist.ChannelDistManager.GetShardLeader(channel.GetChannelName(), replica)
if leader == nil || (!withUnserviceableShards && !leader.IsServiceable()) {
log.WithRateGroup("util.GetShardLeaders", 1, 60).
Warn("leader is not available in replica", zap.String("channel", channel.GetChannelName()), zap.Int64("replicaID", replica.GetID()))
continue
}
info := nodeMgr.Get(leader.Node)
if info != nil {
ids = append(ids, info.ID())
addrs = append(addrs, info.Addr())
serviceable = append(serviceable, leader.IsServiceable())
}
}
if len(ids) == 0 && !withUnserviceableShards {
err := merr.WrapErrChannelNotAvailable(channel.GetChannelName())
msg := fmt.Sprintf("channel %s is not available in any replica", channel.GetChannelName())
log.Warn(msg, zap.Error(err))
return nil, err
}
ret = append(ret, &querypb.ShardLeadersList{
ChannelName: channel.GetChannelName(),
NodeIds: ids,
NodeAddrs: addrs,
Serviceable: serviceable,
})
}
return ret, nil
}
func GetShardLeaders(ctx context.Context,
m *meta.Meta,
targetMgr meta.TargetManagerInterface,
dist *meta.DistributionManager,
nodeMgr *session.NodeManager,
collectionID int64,
withUnserviceableShards bool,
) ([]*querypb.ShardLeadersList, error) {
// skip check load status if withUnserviceableShards is true
if err := checkLoadStatus(ctx, m, collectionID); err != nil {
return nil, err
}
channels := targetMgr.GetDmChannelsByCollection(ctx, collectionID, meta.CurrentTarget)
if len(channels) == 0 {
msg := "loaded collection do not found any channel in target, may be in recovery"
err := merr.WrapErrCollectionOnRecovering(collectionID, msg)
log.Ctx(ctx).Warn("failed to get channels", zap.Error(err))
return nil, err
}
return GetShardLeadersWithChannels(ctx, m, dist, nodeMgr, collectionID, channels, withUnserviceableShards)
}
// CheckCollectionsQueryable check all channels are watched and all segments are loaded for this collection
func CheckCollectionsQueryable(ctx context.Context, m *meta.Meta, targetMgr meta.TargetManagerInterface, dist *meta.DistributionManager, nodeMgr *session.NodeManager) error {
maxInterval := paramtable.Get().QueryCoordCfg.UpdateCollectionLoadStatusInterval.GetAsDuration(time.Minute)
for _, coll := range m.GetAllCollections(ctx) {
err := checkCollectionQueryable(ctx, m, targetMgr, dist, nodeMgr, coll)
// the collection is not queryable, if meet following conditions:
// 1. Some segments are not loaded
// 2. Collection is not starting to release
// 3. The load percentage has not been updated in the last 5 minutes.
if err != nil && m.Exist(ctx, coll.CollectionID) && time.Since(coll.UpdatedAt) >= maxInterval {
log.Ctx(ctx).Warn("collection not querable",
zap.Int64("collectionID", coll.CollectionID),
zap.Time("lastUpdated", coll.UpdatedAt),
zap.Duration("maxInterval", maxInterval),
zap.Error(err))
return err
}
}
return nil
}
// checkCollectionQueryable check all channels are watched and all segments are loaded for this collection
func checkCollectionQueryable(ctx context.Context, m *meta.Meta, targetMgr meta.TargetManagerInterface, dist *meta.DistributionManager, nodeMgr *session.NodeManager, coll *meta.Collection) error {
collectionID := coll.GetCollectionID()
if err := checkLoadStatus(ctx, m, collectionID); err != nil {
return err
}
channels := targetMgr.GetDmChannelsByCollection(ctx, collectionID, meta.CurrentTarget)
if len(channels) == 0 {
msg := "loaded collection do not found any channel in target, may be in recovery"
err := merr.WrapErrCollectionOnRecovering(collectionID, msg)
log.Ctx(ctx).Warn("failed to get channels", zap.Error(err))
return err
}
shardList, err := GetShardLeadersWithChannels(ctx, m, dist, nodeMgr, collectionID, channels, false)
if err != nil {
return err
}
if len(channels) != len(shardList) {
return merr.WrapErrCollectionNotFullyLoaded(collectionID, "still have unwatched channels or loaded segments")
}
return nil
}
func filterDupLeaders(ctx context.Context, 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(ctx, 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
}