milvus/internal/datanode/data_node.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 datanode implements data persistence logic.
//
// Data node persists insert logs into persistent storage like minIO/S3.
package datanode

import (
	"context"
	"fmt"
	"io"
	"math/rand"
	"os"
	"path"
	"strings"
	"sync"
	"sync/atomic"
	"syscall"
	"time"

	"github.com/cockroachdb/errors"
	"github.com/golang/protobuf/proto"
	v3rpc "go.etcd.io/etcd/api/v3/v3rpc/rpctypes"
	clientv3 "go.etcd.io/etcd/client/v3"
	"go.uber.org/zap"

	"github.com/milvus-io/milvus-proto/go-api/v2/commonpb"
	"github.com/milvus-io/milvus/internal/datanode/allocator"
	"github.com/milvus-io/milvus/internal/kv"
	etcdkv "github.com/milvus-io/milvus/internal/kv/etcd"
	"github.com/milvus-io/milvus/internal/proto/datapb"
	"github.com/milvus-io/milvus/internal/proto/rootcoordpb"
	"github.com/milvus-io/milvus/internal/storage"
	"github.com/milvus-io/milvus/internal/types"
	"github.com/milvus-io/milvus/internal/util/dependency"
	"github.com/milvus-io/milvus/internal/util/sessionutil"
	"github.com/milvus-io/milvus/pkg/log"
	"github.com/milvus-io/milvus/pkg/mq/msgdispatcher"
	"github.com/milvus-io/milvus/pkg/util/commonpbutil"
	"github.com/milvus-io/milvus/pkg/util/logutil"
	"github.com/milvus-io/milvus/pkg/util/metricsinfo"
	"github.com/milvus-io/milvus/pkg/util/paramtable"
	"github.com/milvus-io/milvus/pkg/util/retry"
	"github.com/milvus-io/milvus/pkg/util/typeutil"
)

const (
	// RPCConnectionTimeout is used to set the timeout for rpc request
	RPCConnectionTimeout = 30 * time.Second

	// MetricRequestsTotal is used to count the num of total requests
	MetricRequestsTotal = "total"

	// MetricRequestsSuccess is used to count the num of successful requests
	MetricRequestsSuccess = "success"

	// ConnectEtcdMaxRetryTime is used to limit the max retry time for connection etcd
	ConnectEtcdMaxRetryTime = 100

	// ImportCallTimeout is the timeout used in Import() method calls
	// This value is equal to RootCoord's task expire time
	ImportCallTimeout = 15 * 60 * time.Second
)

var getFlowGraphServiceAttempts = uint(50)

// makes sure DataNode implements types.DataNode
var _ types.DataNode = (*DataNode)(nil)

// Params from config.yaml
var Params *paramtable.ComponentParam = paramtable.Get()

// DataNode communicates with outside services and unioun all
// services in datanode package.
//
// DataNode implements `types.Component`, `types.DataNode` interfaces.
//
//	`etcdCli`   is a connection of etcd
//	`rootCoord` is a grpc client of root coordinator.
//	`dataCoord` is a grpc client of data service.
//	`NodeID` is unique to each datanode.
//	`State` is current statement of this data node, indicating whether it's healthy.
//
//	`clearSignal` is a signal channel for releasing the flowgraph resources.
//	`segmentCache` stores all flushing and flushed segments.
type DataNode struct {
	ctx              context.Context
	cancel           context.CancelFunc
	Role             string
	stateCode        atomic.Value // commonpb.StateCode_Initializing
	flowgraphManager *flowgraphManager
	eventManagerMap  *typeutil.ConcurrentMap[string, *channelEventManager]

	clearSignal        chan string // vchannel name
	segmentCache       *Cache
	compactionExecutor *compactionExecutor
	timeTickSender     *timeTickSender

	etcdCli   *clientv3.Client
	address   string
	rootCoord types.RootCoord
	dataCoord types.DataCoord

	//call once
	initOnce     sync.Once
	startOnce    sync.Once
	wg           sync.WaitGroup
	sessionMu    sync.Mutex // to fix data race
	session      *sessionutil.Session
	watchKv      kv.WatchKV
	chunkManager storage.ChunkManager
	allocator    allocator.Allocator

	closer io.Closer

	dispClient msgdispatcher.Client
	factory    dependency.Factory
}

// NewDataNode will return a DataNode with abnormal state.
func NewDataNode(ctx context.Context, factory dependency.Factory) *DataNode {
	rand.Seed(time.Now().UnixNano())
	ctx2, cancel2 := context.WithCancel(ctx)
	node := &DataNode{
		ctx:    ctx2,
		cancel: cancel2,
		Role:   typeutil.DataNodeRole,

		rootCoord:          nil,
		dataCoord:          nil,
		factory:            factory,
		segmentCache:       newCache(),
		compactionExecutor: newCompactionExecutor(),

		eventManagerMap:  typeutil.NewConcurrentMap[string, *channelEventManager](),
		flowgraphManager: newFlowgraphManager(),
		clearSignal:      make(chan string, 100),
	}
	node.UpdateStateCode(commonpb.StateCode_Abnormal)
	return node
}

func (node *DataNode) SetAddress(address string) {
	node.address = address
}

func (node *DataNode) GetAddress() string {
	return node.address
}

// SetEtcdClient sets etcd client for DataNode
func (node *DataNode) SetEtcdClient(etcdCli *clientv3.Client) {
	node.etcdCli = etcdCli
}

// SetRootCoord sets RootCoord's grpc client, error is returned if repeatedly set.
func (node *DataNode) SetRootCoord(rc types.RootCoord) error {
	switch {
	case rc == nil, node.rootCoord != nil:
		return errors.New("nil parameter or repeatedly set")
	default:
		node.rootCoord = rc
		return nil
	}
}

// SetDataCoord sets data service's grpc client, error is returned if repeatedly set.
func (node *DataNode) SetDataCoord(ds types.DataCoord) error {
	switch {
	case ds == nil, node.dataCoord != nil:
		return errors.New("nil parameter or repeatedly set")
	default:
		node.dataCoord = ds
		return nil
	}
}

// Register register datanode to etcd
func (node *DataNode) Register() error {
	node.session.Register()

	// Start liveness check
	node.session.LivenessCheck(node.ctx, func() {
		log.Error("Data Node disconnected from etcd, process will exit", zap.Int64("Server Id", node.GetSession().ServerID))
		if err := node.Stop(); err != nil {
			log.Fatal("failed to stop server", zap.Error(err))
		}
		// manually send signal to starter goroutine
		if node.session.TriggerKill {
			if p, err := os.FindProcess(os.Getpid()); err == nil {
				p.Signal(syscall.SIGINT)
			}
		}
	})

	return nil
}

func (node *DataNode) initSession() error {
	node.session = sessionutil.NewSession(node.ctx, Params.EtcdCfg.MetaRootPath.GetValue(), node.etcdCli)
	if node.session == nil {
		return errors.New("failed to initialize session")
	}
	node.session.Init(typeutil.DataNodeRole, node.address, false, true)
	return nil
}

// initRateCollector creates and starts rateCollector in QueryNode.
func (node *DataNode) initRateCollector() error {
	err := initGlobalRateCollector()
	if err != nil {
		return err
	}
	rateCol.Register(metricsinfo.InsertConsumeThroughput)
	rateCol.Register(metricsinfo.DeleteConsumeThroughput)
	return nil
}

func (node *DataNode) Init() error {
	var initError error
	node.initOnce.Do(func() {
		logutil.Logger(node.ctx).Info("DataNode server initializing",
			zap.String("TimeTickChannelName", Params.CommonCfg.DataCoordTimeTick.GetValue()),
		)
		if err := node.initSession(); err != nil {
			log.Error("DataNode server init session failed", zap.Error(err))
			initError = err
			return
		}

		err := node.initRateCollector()
		if err != nil {
			log.Error("DataNode server init rateCollector failed", zap.Int64("node ID", paramtable.GetNodeID()), zap.Error(err))
			initError = err
			return
		}
		log.Info("DataNode server init rateCollector done", zap.Int64("node ID", paramtable.GetNodeID()))

		node.dispClient = msgdispatcher.NewClient(node.factory, typeutil.DataNodeRole, paramtable.GetNodeID())
		log.Info("DataNode server init dispatcher client done", zap.Int64("node ID", paramtable.GetNodeID()))

		alloc, err := allocator.New(context.Background(), node.rootCoord, paramtable.GetNodeID())
		if err != nil {
			log.Error("failed to create id allocator",
				zap.Error(err),
				zap.String("role", typeutil.DataNodeRole), zap.Int64("DataNode ID", paramtable.GetNodeID()))
			initError = err
			return
		}
		node.allocator = alloc

		node.factory.Init(Params)
		log.Info("DataNode server init succeeded",
			zap.String("MsgChannelSubName", Params.CommonCfg.DataNodeSubName.GetValue()))

	})
	return initError
}

// StartWatchChannels start loop to watch channel allocation status via kv(etcd for now)
func (node *DataNode) StartWatchChannels(ctx context.Context) {
	defer node.wg.Done()
	defer logutil.LogPanic()
	// REF MEP#7 watch path should be [prefix]/channel/{node_id}/{channel_name}
	// TODO, this is risky, we'd better watch etcd with revision rather simply a path
	watchPrefix := path.Join(Params.CommonCfg.DataCoordWatchSubPath.GetValue(), fmt.Sprintf("%d", node.GetSession().ServerID))
	log.Info("Start watch channel", zap.String("prefix", watchPrefix))
	evtChan := node.watchKv.WatchWithPrefix(watchPrefix)
	// after watch, first check all exists nodes first
	err := node.checkWatchedList()
	if err != nil {
		log.Warn("StartWatchChannels failed", zap.Error(err))
		return
	}
	for {
		select {
		case <-ctx.Done():
			log.Info("watch etcd loop quit")
			return
		case event, ok := <-evtChan:
			if !ok {
				log.Warn("datanode failed to watch channel, return")
				go node.StartWatchChannels(ctx)
				return
			}

			if err := event.Err(); err != nil {
				log.Warn("datanode watch channel canceled", zap.Error(event.Err()))
				// https://github.com/etcd-io/etcd/issues/8980
				if event.Err() == v3rpc.ErrCompacted {
					go node.StartWatchChannels(ctx)
					return
				}
				// if watch loop return due to event canceled, the datanode is not functional anymore
				log.Panic("datanode is not functional for event canceled", zap.Error(err))
				return
			}
			for _, evt := range event.Events {
				// We need to stay in order until events enqueued
				node.handleChannelEvt(evt)
			}
		}
	}
}

// checkWatchedList list all nodes under [prefix]/channel/{node_id} and make sure all nodeds are watched
// serves the corner case for etcd connection lost and missing some events
func (node *DataNode) checkWatchedList() error {
	// REF MEP#7 watch path should be [prefix]/channel/{node_id}/{channel_name}
	prefix := path.Join(Params.CommonCfg.DataCoordWatchSubPath.GetValue(), fmt.Sprintf("%d", paramtable.GetNodeID()))
	keys, values, err := node.watchKv.LoadWithPrefix(prefix)
	if err != nil {
		return err
	}
	for i, val := range values {
		node.handleWatchInfo(&event{eventType: putEventType}, keys[i], []byte(val))
	}
	return nil
}

// handleChannelEvt handles event from kv watch event
func (node *DataNode) handleChannelEvt(evt *clientv3.Event) {
	var e *event
	switch evt.Type {
	case clientv3.EventTypePut: // datacoord shall put channels needs to be watched here
		e = &event{
			eventType: putEventType,
			version:   evt.Kv.Version,
		}

	case clientv3.EventTypeDelete:
		e = &event{
			eventType: deleteEventType,
			version:   evt.Kv.Version,
		}
	}
	node.handleWatchInfo(e, string(evt.Kv.Key), evt.Kv.Value)
}

func (node *DataNode) handleWatchInfo(e *event, key string, data []byte) {
	switch e.eventType {
	case putEventType:
		watchInfo, err := parsePutEventData(data)
		if err != nil {
			log.Warn("fail to handle watchInfo", zap.Int("event type", e.eventType), zap.String("key", key), zap.Error(err))
			return
		}

		if isEndWatchState(watchInfo.State) {
			log.Info("DataNode received a PUT event with an end State", zap.String("state", watchInfo.State.String()))
			return
		}

		if watchInfo.Progress != 0 {
			log.Info("DataNode received a PUT event with tickler update progress", zap.String("channel", watchInfo.Vchan.ChannelName), zap.Int64("version", e.version))
			return
		}

		e.info = watchInfo
		e.vChanName = watchInfo.GetVchan().GetChannelName()
		log.Info("DataNode is handling watchInfo PUT event", zap.String("key", key), zap.Any("watch state", watchInfo.GetState().String()))
	case deleteEventType:
		e.vChanName = parseDeleteEventKey(key)
		log.Info("DataNode is handling watchInfo DELETE event", zap.String("key", key))
	}

	actualManager, loaded := node.eventManagerMap.GetOrInsert(e.vChanName, newChannelEventManager(
		node.handlePutEvent, node.handleDeleteEvent, retryWatchInterval,
	))

	if !loaded {
		actualManager.Run()
	}

	actualManager.handleEvent(*e)

	// Whenever a delete event comes, this eventManager will be removed from map
	if e.eventType == deleteEventType {
		if m, loaded := node.eventManagerMap.GetAndRemove(e.vChanName); loaded {
			m.Close()
		}
	}
}

func parsePutEventData(data []byte) (*datapb.ChannelWatchInfo, error) {
	watchInfo := datapb.ChannelWatchInfo{}
	err := proto.Unmarshal(data, &watchInfo)
	if err != nil {
		return nil, fmt.Errorf("invalid event data: fail to parse ChannelWatchInfo, err: %v", err)
	}

	if watchInfo.Vchan == nil {
		return nil, fmt.Errorf("invalid event: ChannelWatchInfo with nil VChannelInfo")
	}
	reviseVChannelInfo(watchInfo.GetVchan())
	return &watchInfo, nil
}

func parseDeleteEventKey(key string) string {
	parts := strings.Split(key, "/")
	vChanName := parts[len(parts)-1]
	return vChanName
}

func (node *DataNode) handlePutEvent(watchInfo *datapb.ChannelWatchInfo, version int64) (err error) {
	vChanName := watchInfo.GetVchan().GetChannelName()
	key := path.Join(Params.CommonCfg.DataCoordWatchSubPath.GetValue(), fmt.Sprintf("%d", node.GetSession().ServerID), vChanName)
	tickler := newTickler(version, key, watchInfo, node.watchKv, Params.DataNodeCfg.WatchEventTicklerInterval.GetAsDuration(time.Second))

	switch watchInfo.State {
	case datapb.ChannelWatchState_Uncomplete, datapb.ChannelWatchState_ToWatch:
		if err := node.flowgraphManager.addAndStart(node, watchInfo.GetVchan(), watchInfo.GetSchema(), tickler); err != nil {
			watchInfo.State = datapb.ChannelWatchState_WatchFailure
			return fmt.Errorf("fail to add and start flowgraph for vChanName: %s, err: %v", vChanName, err)
		}
		log.Info("handle put event: new data sync service success", zap.String("vChanName", vChanName))
		watchInfo.State = datapb.ChannelWatchState_WatchSuccess

	case datapb.ChannelWatchState_ToRelease:
		// there is no reason why we release fail
		node.tryToReleaseFlowgraph(vChanName)
		watchInfo.State = datapb.ChannelWatchState_ReleaseSuccess
	}

	v, err := proto.Marshal(watchInfo)
	if err != nil {
		return fmt.Errorf("fail to marshal watchInfo with state, vChanName: %s, state: %s ,err: %w", vChanName, watchInfo.State.String(), err)
	}

	success, err := node.watchKv.CompareVersionAndSwap(key, tickler.version, string(v))
	// etcd error
	if err != nil {
		// flow graph will leak if not release, causing new datanode failed to subscribe
		node.tryToReleaseFlowgraph(vChanName)
		log.Warn("fail to update watch state to etcd", zap.String("vChanName", vChanName),
			zap.String("state", watchInfo.State.String()), zap.Error(err))
		return err
	}
	// etcd valid but the states updated.
	if !success {
		log.Info("handle put event: failed to compare version and swap, release flowgraph",
			zap.String("key", key), zap.String("state", watchInfo.State.String()),
			zap.String("vChanName", vChanName))
		// flow graph will leak if not release, causing new datanode failed to subscribe
		node.tryToReleaseFlowgraph(vChanName)
		return nil
	}
	log.Info("handle put event success", zap.String("key", key),
		zap.String("state", watchInfo.State.String()), zap.String("vChanName", vChanName))
	return nil
}

func (node *DataNode) handleDeleteEvent(vChanName string) {
	node.tryToReleaseFlowgraph(vChanName)
}

// tryToReleaseFlowgraph tries to release a flowgraph
func (node *DataNode) tryToReleaseFlowgraph(vChanName string) {
	log.Info("try to release flowgraph", zap.String("vChanName", vChanName))
	node.flowgraphManager.release(vChanName)
}

// BackGroundGC runs in background to release datanode resources
// GOOSE TODO: remove background GC, using ToRelease for drop-collection after #15846
func (node *DataNode) BackGroundGC(vChannelCh <-chan string) {
	defer node.wg.Done()
	log.Info("DataNode Background GC Start")
	for {
		select {
		case vchanName := <-vChannelCh:
			node.tryToReleaseFlowgraph(vchanName)
		case <-node.ctx.Done():
			log.Warn("DataNode context done, exiting background GC")
			return
		}
	}
}

// Start will update DataNode state to HEALTHY
func (node *DataNode) Start() error {
	var startErr error
	node.startOnce.Do(func() {
		if err := node.allocator.Start(); err != nil {
			log.Error("failed to start id allocator", zap.Error(err), zap.String("role", typeutil.DataNodeRole))
			startErr = err
			return
		}
		log.Info("start id allocator done", zap.String("role", typeutil.DataNodeRole))

		rep, err := node.rootCoord.AllocTimestamp(node.ctx, &rootcoordpb.AllocTimestampRequest{
			Base: commonpbutil.NewMsgBase(
				commonpbutil.WithMsgType(commonpb.MsgType_RequestTSO),
				commonpbutil.WithMsgID(0),
				commonpbutil.WithSourceID(paramtable.GetNodeID()),
			),
			Count: 1,
		})
		if err != nil || rep.Status.ErrorCode != commonpb.ErrorCode_Success {
			log.Warn("fail to alloc timestamp", zap.Any("rep", rep), zap.Error(err))
			startErr = errors.New("DataNode fail to alloc timestamp")
			return
		}

		connectEtcdFn := func() error {
			etcdKV := etcdkv.NewEtcdKV(node.etcdCli, Params.EtcdCfg.MetaRootPath.GetValue())
			node.watchKv = etcdKV
			return nil
		}
		err = retry.Do(node.ctx, connectEtcdFn, retry.Attempts(ConnectEtcdMaxRetryTime))
		if err != nil {
			startErr = errors.New("DataNode fail to connect etcd")
			return
		}

		chunkManager, err := node.factory.NewPersistentStorageChunkManager(node.ctx)

		if err != nil {
			startErr = err
			return
		}

		node.chunkManager = chunkManager

		node.wg.Add(1)
		go node.BackGroundGC(node.clearSignal)

		go node.compactionExecutor.start(node.ctx)

		if Params.DataNodeCfg.DataNodeTimeTickByRPC.GetAsBool() {
			node.timeTickSender = newTimeTickManager(node.dataCoord, node.session.ServerID)
			go node.timeTickSender.start(node.ctx)
		}

		node.wg.Add(1)
		// Start node watch node
		go node.StartWatchChannels(node.ctx)

		go node.flowgraphManager.start()

		node.UpdateStateCode(commonpb.StateCode_Healthy)

	})
	return startErr
}

// UpdateStateCode updates datanode's state code
func (node *DataNode) UpdateStateCode(code commonpb.StateCode) {
	node.stateCode.Store(code)
}

// GetStateCode return datanode's state code
func (node *DataNode) GetStateCode() commonpb.StateCode {
	return node.stateCode.Load().(commonpb.StateCode)
}

func (node *DataNode) isHealthy() bool {
	return node.GetStateCode() == commonpb.StateCode_Healthy
}

// ReadyToFlush tells whether DataNode is ready for flushing
func (node *DataNode) ReadyToFlush() error {
	if !node.isHealthy() {
		return errors.New("DataNode not in HEALTHY state")
	}
	return nil
}

// Stop will release DataNode resources and shutdown datanode
func (node *DataNode) Stop() error {
	// https://github.com/milvus-io/milvus/issues/12282
	node.UpdateStateCode(commonpb.StateCode_Abnormal)
	node.flowgraphManager.dropAll()
	node.flowgraphManager.stop()

	node.cancel()
	node.wg.Wait()
	node.eventManagerMap.Range(func(_ string, m *channelEventManager) bool {
		m.Close()
		return true
	})

	if node.allocator != nil {
		log.Info("close id allocator", zap.String("role", typeutil.DataNodeRole))
		node.allocator.Close()
	}

	if node.closer != nil {
		err := node.closer.Close()
		if err != nil {
			return err
		}
	}

	if node.session != nil {
		node.session.Stop()
	}

	return nil
}

// to fix data race
func (node *DataNode) SetSession(session *sessionutil.Session) {
	node.sessionMu.Lock()
	defer node.sessionMu.Unlock()
	node.session = session
}

// to fix data race
func (node *DataNode) GetSession() *sessionutil.Session {
	node.sessionMu.Lock()
	defer node.sessionMu.Unlock()
	return node.session
}