rondagostino commented on a change in pull request #10069: URL: https://github.com/apache/kafka/pull/10069#discussion_r571237871
########## 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, + highWatermarkCheckpoints, -1, mostRecentMetadataOffsets) + + // We need to transition anything that hasn't transitioned from Deferred to Offline to the Online state. + // We also need to identify the leadership change callback(s) to invoke + deferredPartitionsIterator.foreach { deferredPartition => + val state = deferredPartition.metadata + val partition = deferredPartition.partition + val topicPartition = partition.topicPartition + // identify for callback if necessary + if (state.leaderId == localBrokerId) { + if (partitionsMadeLeader.contains(partition)) { + leadershipChangeCallbacks.getOrElseUpdate( + deferredPartition.onLeadershipChange, (mutable.Set(), mutable.Set()))._1 += partition + } + } else if (partitionsMadeFollower.contains(partition)) { + leadershipChangeCallbacks.getOrElseUpdate( + deferredPartition.onLeadershipChange, (mutable.Set(), mutable.Set()))._2 += partition + } + // transition from Deferred to Online + allPartitions.put(topicPartition, HostedPartition.Online(partition)) + } + + updateLeaderAndFollowerMetrics(partitionsMadeFollower.map(_.topic).toSet) + + maybeAddLogDirFetchers(partitionsMadeFollower, highWatermarkCheckpoints) + + replicaFetcherManager.shutdownIdleFetcherThreads() + replicaAlterLogDirsManager.shutdownIdleFetcherThreads() + leadershipChangeCallbacks.forKeyValue { (onLeadershipChange, leaderAndFollowerPartitions) => + onLeadershipChange(leaderAndFollowerPartitions._1, leaderAndFollowerPartitions._2) + } + } catch { + case e: Throwable => + deferredPartitionsIterator.foreach { metadata => + val state = metadata.metadata + val partition = metadata.partition + val topicPartition = partition.topicPartition + val mostRecentMetadataOffset = metadata.mostRecentMetadataOffset + val leader = state.leaderId == localBrokerId + val leaderOrFollower = if (leader) "leader" else "follower" + val partitionLogMsgPrefix = s"Apply deferred $leaderOrFollower partition $topicPartition last seen in metadata batch $mostRecentMetadataOffset" + stateChangeLogger.error(s"$partitionLogMsgPrefix: error while applying deferred metadata.", e) + } + stateChangeLogger.info(s"Applied ${partitionsMadeLeader.size + partitionsMadeFollower.size} deferred partitions prior to the error: " + + s"${partitionsMadeLeader.size} leader(s) and ${partitionsMadeFollower.size} follower(s)") + // Re-throw the exception for it to be caught in BrokerMetadataListener + throw e Review comment: > If we fail to apply changes, I guess we have to see that as a fatal error? The only possible way of recovering would be to replay the changes. Yeah, I think so. We may need to put effort into minimizing the blast radius of these failures. ########## 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: > Do we have any guarantee that the metadata cache is in a state that is consistent with the deferred changes? Good question. Metadata changes up through the point of applying the deferred partition metadata changes should be applied to the metadata cache at this point. One thing we need to think about is the fact that we currently don't defer metadata cache changes at all. The metadata cache will contain partition states that are ahead of ReplicaManager during the time when ReplicaManager is deferring its changes. This means, for example, that the following will reflect deferred partition changes that have been applied to the metadata cache but that have not been applied to ReplicaManager. We may have to write test cases for each of these conditions so we can be clear on what the expected behavior should be. - MetadataRequest - FindCoordinatorRequest - ElectLeadersRequest with topicPartitions = null - DelayedCreatePartitions (in topic purgatory) - DelayedElectLeader (in elect leader purgatory) - Anything that calls ReplicaManager.fetchMessages() and DelayedFetch (in fetch purgatory), though these seem okay since they wait until they can get enough data? - TransactionMarkerChannelManager.addTxnMarkersToBrokerQueue - DescribeConfigsRequest - OffsetCommitRequest (whether or not to send UNKNOWN_TOPIC_OR_PARTITION) - ProduceRequest (whether or not to send UNKNOWN_TOPIC_OR_PARTITION) - FetchRequest (whether or not to send UNKNOWN_TOPIC_OR_PARTITION) - DeleteRecordsRequest (whether or not to send UNKNOWN_TOPIC_OR_PARTITION) - AddPartitionsToTxnRequest (whether or not to send UNKNOWN_TOPIC_OR_PARTITION) - TxnOffsetCommitRequest (whether or not to send UNKNOWN_TOPIC_OR_PARTITION) - OffsetDeleteRequest (whether or not to send UNKNOWN_TOPIC_OR_PARTITION) ---------------------------------------------------------------- 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]
