rondagostino commented on a change in pull request #10069: URL: https://github.com/apache/kafka/pull/10069#discussion_r571309365
########## File path: core/src/main/scala/kafka/server/RaftReplicaManager.scala ########## @@ -0,0 +1,538 @@ +/** + * Licensed to the Apache Software Foundation (ASF) 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 kafka.server + +import java.util +import java.util.concurrent.atomic.AtomicBoolean + +import kafka.cluster.Partition +import kafka.log.LogManager +import kafka.server.QuotaFactory.QuotaManagers +import kafka.server.checkpoints.{LazyOffsetCheckpoints, OffsetCheckpoints} +import kafka.server.metadata.{ConfigRepository, MetadataBroker, MetadataBrokers, MetadataImageBuilder, MetadataPartition} +import kafka.utils.Implicits.MapExtensionMethods +import kafka.utils.Scheduler +import org.apache.kafka.common.{Node, TopicPartition} +import org.apache.kafka.common.errors.KafkaStorageException +import org.apache.kafka.common.metrics.Metrics +import org.apache.kafka.common.utils.Time + +import scala.collection.{Map, Set, mutable} + +class RaftReplicaManager(config: KafkaConfig, + metrics: Metrics, + time: Time, + scheduler: Scheduler, + logManager: LogManager, + isShuttingDown: AtomicBoolean, + quotaManagers: QuotaManagers, + brokerTopicStats: BrokerTopicStats, + metadataCache: MetadataCache, + logDirFailureChannel: LogDirFailureChannel, + delayedProducePurgatory: DelayedOperationPurgatory[DelayedProduce], + delayedFetchPurgatory: DelayedOperationPurgatory[DelayedFetch], + delayedDeleteRecordsPurgatory: DelayedOperationPurgatory[DelayedDeleteRecords], + delayedElectLeaderPurgatory: DelayedOperationPurgatory[DelayedElectLeader], + threadNamePrefix: Option[String], + configRepository: ConfigRepository, + alterIsrManager: AlterIsrManager) extends ReplicaManager( + config, metrics, time, None, scheduler, logManager, isShuttingDown, quotaManagers, + brokerTopicStats, metadataCache, logDirFailureChannel, delayedProducePurgatory, delayedFetchPurgatory, + delayedDeleteRecordsPurgatory, delayedElectLeaderPurgatory, threadNamePrefix, configRepository, alterIsrManager) { + + if (config.requiresZookeeper) { + throw new IllegalStateException(s"Cannot use ${getClass.getSimpleName} when using ZooKeeper") + } + + // Changes are initially deferred when using a Raft-based metadata quorum, and they may flip-flop to not + // being deferred and being deferred again thereafter as the broker (re)acquires/loses its lease. + // Changes are never deferred when using ZooKeeper. When true, this indicates that we should transition + // online partitions to the deferred state if we see a metadata update for that partition. + private var deferringMetadataChanges: Boolean = true + stateChangeLogger.debug(s"Metadata changes are initially being deferred") + + def beginMetadataChangeDeferral(): Unit = { + replicaStateChangeLock synchronized { + deferringMetadataChanges = true + stateChangeLogger.info(s"Metadata changes are now being deferred") + } + } + + def endMetadataChangeDeferral(): Unit = { + val startMs = time.milliseconds() + replicaStateChangeLock synchronized { + stateChangeLogger.info(s"Applying deferred metadata changes") + val highWatermarkCheckpoints = new LazyOffsetCheckpoints(this.highWatermarkCheckpoints) + val partitionsMadeFollower = mutable.Set[Partition]() + val partitionsMadeLeader = mutable.Set[Partition]() + val leadershipChangeCallbacks = + mutable.Map[(Iterable[Partition], Iterable[Partition]) => Unit, (mutable.Set[Partition], mutable.Set[Partition])]() + try { + val leaderPartitionStates = mutable.Map[Partition, MetadataPartition]() + val followerPartitionStates = mutable.Map[Partition, MetadataPartition]() + val partitionsAlreadyExisting = mutable.Set[MetadataPartition]() + val mostRecentMetadataOffsets = mutable.Map[Partition, Long]() + deferredPartitionsIterator.foreach { deferredPartition => + val state = deferredPartition.metadata + val partition = deferredPartition.partition + if (state.leaderId == localBrokerId) { + leaderPartitionStates.put(partition, state) + } else { + followerPartitionStates.put(partition, state) + } + if (!deferredPartition.isNew) { + partitionsAlreadyExisting += state + } + mostRecentMetadataOffsets.put(partition, deferredPartition.mostRecentMetadataOffset) + } + + val partitionsMadeLeader = makeLeaders(partitionsAlreadyExisting, leaderPartitionStates, + highWatermarkCheckpoints,-1, mostRecentMetadataOffsets) + val partitionsMadeFollower = makeFollowers(partitionsAlreadyExisting, + createMetadataBrokersFromCurrentCache, followerPartitionStates, Review comment: I added the partition state with respect to deferral to `MetadataPartitions`. Storing the information in the metadat cache could help here. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
