tasks/service/history/replication/task_processor_manager.go

// The MIT License
//
// Copyright (c) 2020 Temporal Technologies Inc.  All rights reserved.
//
// Copyright (c) 2020 Uber Technologies, Inc.
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.

package replication

import (
	"context"
	"sync"
	"sync/atomic"
	"time"

	"go.temporal.io/server/api/historyservice/v1"
	persistencespb "go.temporal.io/server/api/persistence/v1"
	"go.temporal.io/server/client"
	"go.temporal.io/server/common"
	"go.temporal.io/server/common/backoff"
	"go.temporal.io/server/common/cluster"
	"go.temporal.io/server/common/headers"
	"go.temporal.io/server/common/log"
	"go.temporal.io/server/common/log/tag"
	"go.temporal.io/server/common/metrics"
	"go.temporal.io/server/common/persistence"
	"go.temporal.io/server/common/persistence/serialization"
	"go.temporal.io/server/common/util"
	"go.temporal.io/server/common/xdc"
	"go.temporal.io/server/service/history/configs"
	"go.temporal.io/server/service/history/deletemanager"
	"go.temporal.io/server/service/history/shard"
	"go.temporal.io/server/service/history/tasks"
	wcache "go.temporal.io/server/service/history/workflow/cache"
)

const (
	dlqSizeCheckInterval = time.Minute * 5
)

type (
	// taskProcessorManagerImpl is to manage replication task processors
	taskProcessorManagerImpl struct {
		config                        *configs.Config
		deleteMgr                     deletemanager.DeleteManager
		engine                        shard.Engine
		eventSerializer               serialization.Serializer
		shard                         shard.Context
		status                        int32
		replicationTaskFetcherFactory TaskFetcherFactory
		workflowCache                 wcache.Cache
		resender                      xdc.NDCHistoryResender
		taskExecutorProvider          TaskExecutorProvider
		taskPollerManager             pollerManager
		metricsHandler                metrics.Handler
		logger                        log.Logger

		enableFetcher     bool
		taskProcessorLock sync.RWMutex
		taskProcessors    map[string][]TaskProcessor // cluster name - processor
		minTxAckedTaskID  int64
		shutdownChan      chan struct{}
	}
)

var _ common.Daemon = (*taskProcessorManagerImpl)(nil)

func NewTaskProcessorManager(
	config *configs.Config,
	shard shard.Context,
	engine shard.Engine,
	workflowCache wcache.Cache,
	workflowDeleteManager deletemanager.DeleteManager,
	clientBean client.Bean,
	eventSerializer serialization.Serializer,
	replicationTaskFetcherFactory TaskFetcherFactory,
	taskExecutorProvider TaskExecutorProvider,
) *taskProcessorManagerImpl {

	return &taskProcessorManagerImpl{
		config:                        config,
		deleteMgr:                     workflowDeleteManager,
		engine:                        engine,
		eventSerializer:               eventSerializer,
		shard:                         shard,
		status:                        common.DaemonStatusInitialized,
		replicationTaskFetcherFactory: replicationTaskFetcherFactory,
		workflowCache:                 workflowCache,
		resender: xdc.NewNDCHistoryResender(
			shard.GetNamespaceRegistry(),
			clientBean,
			func(ctx context.Context, request *historyservice.ReplicateEventsV2Request) error {
				_, err := clientBean.GetHistoryClient().ReplicateEventsV2(ctx, request)
				return err
			},
			shard.GetPayloadSerializer(),
			shard.GetConfig().StandbyTaskReReplicationContextTimeout,
			shard.GetLogger(),
		),
		logger:         shard.GetLogger(),
		metricsHandler: shard.GetMetricsHandler(),

		enableFetcher:        !config.EnableReplicationStream(),
		taskProcessors:       make(map[string][]TaskProcessor),
		taskExecutorProvider: taskExecutorProvider,
		taskPollerManager:    newPollerManager(shard.GetShardID(), shard.GetClusterMetadata()),
		minTxAckedTaskID:     persistence.EmptyQueueMessageID,
		shutdownChan:         make(chan struct{}),
	}
}

func (r *taskProcessorManagerImpl) Start() {
	if !atomic.CompareAndSwapInt32(
		&r.status,
		common.DaemonStatusInitialized,
		common.DaemonStatusStarted,
	) {
		return
	}

	// Listen to cluster metadata and dynamically update replication processor for remote clusters.
	if r.enableFetcher {
		r.listenToClusterMetadataChange()
	}
	go r.completeReplicationTaskLoop()
	go r.checkReplicationDLQEmptyLoop()
}

func (r *taskProcessorManagerImpl) Stop() {
	if !atomic.CompareAndSwapInt32(
		&r.status,
		common.DaemonStatusStarted,
		common.DaemonStatusStopped,
	) {
		return
	}

	close(r.shutdownChan)

	if r.enableFetcher {
		r.shard.GetClusterMetadata().UnRegisterMetadataChangeCallback(r)
	}
	r.taskProcessorLock.Lock()
	for _, taskProcessors := range r.taskProcessors {
		for _, processor := range taskProcessors {
			processor.Stop()
		}
	}
	r.taskProcessorLock.Unlock()
}

func (r *taskProcessorManagerImpl) listenToClusterMetadataChange() {
	clusterMetadata := r.shard.GetClusterMetadata()
	clusterMetadata.RegisterMetadataChangeCallback(
		r,
		r.handleClusterMetadataUpdate,
	)
}

func (r *taskProcessorManagerImpl) handleClusterMetadataUpdate(
	oldClusterMetadata map[string]*cluster.ClusterInformation,
	newClusterMetadata map[string]*cluster.ClusterInformation,
) {
	r.taskProcessorLock.Lock()
	defer r.taskProcessorLock.Unlock()
	currentClusterName := r.shard.GetClusterMetadata().GetCurrentClusterName()
	// The metadata triggers an update when the following fields update: 1. Enabled 2. Initial Failover Version 3. Cluster address
	// The callback covers three cases:
	// Case 1: Remove a cluster Case 2: Add a new cluster Case 3: Refresh cluster metadata(1 + 2).

	// Case 1 and Case 3
	for clusterName := range oldClusterMetadata {
		if clusterName == currentClusterName {
			continue
		}
		for _, processor := range r.taskProcessors[clusterName] {
			processor.Stop()
		}
		delete(r.taskProcessors, clusterName)
	}

	// Case 2 and Case 3
	for clusterName := range newClusterMetadata {
		if clusterName == currentClusterName {
			continue
		}
		if clusterInfo := newClusterMetadata[clusterName]; clusterInfo == nil || !clusterInfo.Enabled {
			continue
		}
		sourceShardIds, err := r.taskPollerManager.getSourceClusterShardIDs(clusterName)
		if err != nil {
			r.logger.Error("Failed to get source shard id list", tag.Error(err), tag.ClusterName(clusterName))
			continue
		}
		var processors []TaskProcessor
		for _, sourceShardId := range sourceShardIds {
			fetcher := r.replicationTaskFetcherFactory.GetOrCreateFetcher(clusterName)
			replicationTaskProcessor := NewTaskProcessor(
				sourceShardId,
				r.shard,
				r.engine,
				r.config,
				r.shard.GetMetricsHandler(),
				fetcher,
				r.taskExecutorProvider(TaskExecutorParams{
					RemoteCluster:   clusterName,
					Shard:           r.shard,
					HistoryResender: r.resender,
					DeleteManager:   r.deleteMgr,
					WorkflowCache:   r.workflowCache,
				}),
				r.eventSerializer,
			)
			replicationTaskProcessor.Start()
			processors = append(processors, replicationTaskProcessor)
		}
		r.taskProcessors[clusterName] = processors
	}
}

func (r *taskProcessorManagerImpl) completeReplicationTaskLoop() {
	shardID := r.shard.GetShardID()
	cleanupTimer := time.NewTimer(backoff.Jitter(
		r.config.ReplicationTaskProcessorCleanupInterval(shardID),
		r.config.ReplicationTaskProcessorCleanupJitterCoefficient(shardID),
	))
	defer cleanupTimer.Stop()
	for {
		select {
		case <-cleanupTimer.C:
			if err := r.cleanupReplicationTasks(); err != nil {
				r.logger.Error("Failed to clean up replication messages.", tag.Error(err))
				r.metricsHandler.Counter(metrics.ReplicationTaskCleanupFailure.GetMetricName()).Record(
					1,
					metrics.OperationTag(metrics.ReplicationTaskCleanupScope),
				)
			}
			cleanupTimer.Reset(backoff.Jitter(
				r.config.ReplicationTaskProcessorCleanupInterval(shardID),
				r.config.ReplicationTaskProcessorCleanupJitterCoefficient(shardID),
			))
		case <-r.shutdownChan:
			return
		}
	}
}

func (r *taskProcessorManagerImpl) checkReplicationDLQEmptyLoop() {
	for {
		select {
		case <-time.After(backoff.FullJitter(dlqSizeCheckInterval)):
			if r.config.ReplicationEnableDLQMetrics() {
				r.checkReplicationDLQSize()
			}
		case <-r.shutdownChan:
			return
		}
	}
}

func (r *taskProcessorManagerImpl) cleanupReplicationTasks() error {
	clusterMetadata := r.shard.GetClusterMetadata()
	allClusterInfo := clusterMetadata.GetAllClusterInfo()
	currentClusterName := clusterMetadata.GetCurrentClusterName()

	minAckedTaskID := r.shard.GetImmediateQueueExclusiveHighReadWatermark().TaskID - 1
	queueStates, ok := r.shard.GetQueueState(tasks.CategoryReplication)
	if !ok {
		queueStates = &persistencespb.QueueState{
			ExclusiveReaderHighWatermark: nil,
			ReaderStates:                 make(map[int64]*persistencespb.QueueReaderState),
		}
	}
	for _, readerID := range targetReaderIDs(
		currentClusterName,
		r.shard.GetShardID(),
		allClusterInfo,
	) {
		readerState, ok := queueStates.ReaderStates[readerID]
		if !ok {
			minAckedTaskID = util.Min[int64](minAckedTaskID, 0)
		} else {
			minAckedTaskID = util.Min[int64](minAckedTaskID, readerState.Scopes[0].Range.InclusiveMin.TaskId-1)
		}
	}

	if minAckedTaskID <= r.minTxAckedTaskID {
		return nil
	}

	r.logger.Debug("cleaning up replication task queue", tag.ReadLevel(minAckedTaskID))
	r.metricsHandler.Counter(metrics.ReplicationTaskCleanupCount.GetMetricName()).Record(
		1,
		metrics.OperationTag(metrics.ReplicationTaskCleanupScope),
	)
	r.metricsHandler.Histogram(metrics.ReplicationTasksLag.GetMetricName(), metrics.ReplicationTasksLag.GetMetricUnit()).Record(
		r.shard.GetImmediateQueueExclusiveHighReadWatermark().Prev().TaskID-minAckedTaskID,
		metrics.TargetClusterTag(currentClusterName),
		metrics.OperationTag(metrics.ReplicationTaskCleanupScope),
	)

	ctx, cancel := context.WithTimeout(context.Background(), time.Second*5)
	ctx = headers.SetCallerInfo(ctx, headers.SystemPreemptableCallerInfo)
	defer cancel()

	inclusiveMinPendingTaskKey := tasks.NewImmediateKey(minAckedTaskID + 1)
	r.shard.GetExecutionManager().UpdateHistoryTaskReaderProgress(ctx, &persistence.UpdateHistoryTaskReaderProgressRequest{
		ShardID:                    r.shard.GetShardID(),
		ShardOwner:                 r.shard.GetOwner(),
		TaskCategory:               tasks.CategoryReplication,
		ReaderID:                   common.DefaultQueueReaderID,
		InclusiveMinPendingTaskKey: inclusiveMinPendingTaskKey,
	})

	err := r.shard.GetExecutionManager().RangeCompleteHistoryTasks(
		ctx,
		&persistence.RangeCompleteHistoryTasksRequest{
			ShardID:             r.shard.GetShardID(),
			TaskCategory:        tasks.CategoryReplication,
			ExclusiveMaxTaskKey: inclusiveMinPendingTaskKey,
		},
	)
	if err == nil {
		r.minTxAckedTaskID = minAckedTaskID
	}
	return err
}

func (r *taskProcessorManagerImpl) checkReplicationDLQSize() {
	ctx, cancel := context.WithTimeout(context.Background(), time.Second*5)
	ctx = headers.SetCallerInfo(ctx, headers.SystemPreemptableCallerInfo)
	defer cancel()
	for clusterName := range r.shard.GetClusterMetadata().GetAllClusterInfo() {
		if clusterName == r.shard.GetClusterMetadata().GetCurrentClusterName() {
			continue
		}

		minTaskKey := r.shard.GetReplicatorDLQAckLevel(clusterName)
		isEmpty, err := r.shard.GetExecutionManager().IsReplicationDLQEmpty(ctx, &persistence.GetReplicationTasksFromDLQRequest{
			GetHistoryTasksRequest: persistence.GetHistoryTasksRequest{
				ShardID:             r.shard.GetShardID(),
				TaskCategory:        tasks.CategoryReplication,
				InclusiveMinTaskKey: tasks.NewImmediateKey(minTaskKey),
			},
			SourceClusterName: clusterName,
		})
		if err != nil {
			r.logger.Error("Failed to check replication DLQ size.", tag.Error(err))
			return
		}
		if !isEmpty {
			r.metricsHandler.Counter(metrics.ReplicationNonEmptyDLQCount.GetMetricName()).Record(1, metrics.OperationTag(metrics.ReplicationDLQStatsScope))
			r.logger.Info("Replication DLQ is not empty.", tag.ShardID(r.shard.GetShardID()), tag.AckLevel(minTaskKey))
			break
		}
	}
}

func targetReaderIDs(
	currentClusterName string,
	currentShardID int32,
	allClusterInfo map[string]cluster.ClusterInformation,
) []int64 {
	currentShardCount := allClusterInfo[currentClusterName].ShardCount
	var readerIDs []int64
	for clusterName, clusterInfo := range allClusterInfo {
		if clusterName == currentClusterName || !clusterInfo.Enabled {
			continue
		}

		targetClusterID := allClusterInfo[clusterName].InitialFailoverVersion
		targetShardCount := allClusterInfo[clusterName].ShardCount
		for _, targetShardID := range common.MapShardID(
			currentShardCount,
			targetShardCount,
			currentShardID,
		) {
			readerIDs = append(readerIDs, shard.ReplicationReaderIDFromClusterShardID(
				targetClusterID, targetShardID,
			))
		}
	}
	return readerIDs
}