junrao commented on a change in pull request #10579: URL: https://github.com/apache/kafka/pull/10579#discussion_r667102624
########## File path: storage/src/main/java/org/apache/kafka/server/log/remote/metadata/storage/TopicBasedRemoteLogMetadataManager.java ########## @@ -0,0 +1,432 @@ +/* + * 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 org.apache.kafka.server.log.remote.metadata.storage; + +import org.apache.kafka.clients.admin.AdminClient; +import org.apache.kafka.clients.admin.NewTopic; +import org.apache.kafka.clients.producer.RecordMetadata; +import org.apache.kafka.common.KafkaException; +import org.apache.kafka.common.TopicIdPartition; +import org.apache.kafka.common.config.TopicConfig; +import org.apache.kafka.common.errors.RetriableException; +import org.apache.kafka.common.errors.TopicExistsException; +import org.apache.kafka.common.utils.KafkaThread; +import org.apache.kafka.common.utils.Time; +import org.apache.kafka.common.utils.Utils; +import org.apache.kafka.server.log.remote.storage.RemoteLogMetadata; +import org.apache.kafka.server.log.remote.storage.RemoteLogMetadataManager; +import org.apache.kafka.server.log.remote.storage.RemoteLogSegmentMetadata; +import org.apache.kafka.server.log.remote.storage.RemoteLogSegmentMetadataUpdate; +import org.apache.kafka.server.log.remote.storage.RemoteLogSegmentState; +import org.apache.kafka.server.log.remote.storage.RemotePartitionDeleteMetadata; +import org.apache.kafka.server.log.remote.storage.RemoteStorageException; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import java.io.IOException; +import java.time.Duration; +import java.util.Collections; +import java.util.HashMap; +import java.util.HashSet; +import java.util.Iterator; +import java.util.Map; +import java.util.Objects; +import java.util.Optional; +import java.util.Set; +import java.util.concurrent.atomic.AtomicBoolean; +import java.util.concurrent.locks.ReentrantReadWriteLock; + +/** + * This is the {@link RemoteLogMetadataManager} implementation with storage as an internal topic with name {@link TopicBasedRemoteLogMetadataManagerConfig#REMOTE_LOG_METADATA_TOPIC_NAME}. + * This is used to publish and fetch {@link RemoteLogMetadata} for the registered user topic partitions with + * {@link #onPartitionLeadershipChanges(Set, Set)}. Each broker will have an instance of this class and it subscribes + * to metadata updates for the registered user topic partitions. + */ +public class TopicBasedRemoteLogMetadataManager implements RemoteLogMetadataManager { + private static final Logger log = LoggerFactory.getLogger(TopicBasedRemoteLogMetadataManager.class); + private static final long INITIALIZATION_RETRY_INTERVAL_MS = 30000L; + + private volatile boolean configured = false; + + // It indicates whether the close process of this instance is started or not via #close() method. + // Using AtomicBoolean instead of volatile as it may encounter http://findbugs.sourceforge.net/bugDescriptions.html#SP_SPIN_ON_FIELD + // if the field is read but not updated in a spin loop like in #initializeResources() method. + private final AtomicBoolean closing = new AtomicBoolean(false); + private final AtomicBoolean initialized = new AtomicBoolean(false); + private final Time time = Time.SYSTEM; + + private Thread initializationThread; + private volatile ProducerManager producerManager; + private volatile ConsumerManager consumerManager; + + // This allows to gracefully close this instance using {@link #close()} method while there are some pending or new + // requests calling different methods which use the resources like producer/consumer managers. + private final ReentrantReadWriteLock lock = new ReentrantReadWriteLock(); + + private final RemotePartitionMetadataStore remotePartitionMetadataStore = new RemotePartitionMetadataStore(); + private volatile TopicBasedRemoteLogMetadataManagerConfig rlmmConfig; + private volatile RemoteLogMetadataTopicPartitioner rlmmTopicPartitioner; + private volatile Set<TopicIdPartition> pendingAssignPartitions = Collections.synchronizedSet(new HashSet<>()); + + @Override + public void addRemoteLogSegmentMetadata(RemoteLogSegmentMetadata remoteLogSegmentMetadata) + throws RemoteStorageException { + Objects.requireNonNull(remoteLogSegmentMetadata, "remoteLogSegmentMetadata can not be null"); + + // This allows gracefully rejecting the requests while closing of this instance is in progress, which triggers + // closing the producer/consumer manager instances. + lock.readLock().lock(); + try { + ensureInitializedAndNotClosed(); + + // This method is allowed only to add remote log segment with the initial state(which is RemoteLogSegmentState.COPY_SEGMENT_STARTED) + // but not to update the existing remote log segment metadata. + if (remoteLogSegmentMetadata.state() != RemoteLogSegmentState.COPY_SEGMENT_STARTED) { + throw new IllegalArgumentException( + "Given remoteLogSegmentMetadata should have state as " + RemoteLogSegmentState.COPY_SEGMENT_STARTED + + " but it contains state as: " + remoteLogSegmentMetadata.state()); + } + + // Publish the message to the topic. + doPublishMetadata(remoteLogSegmentMetadata.remoteLogSegmentId().topicIdPartition(), + remoteLogSegmentMetadata); + } finally { + lock.readLock().unlock(); + } + } + + @Override + public void updateRemoteLogSegmentMetadata(RemoteLogSegmentMetadataUpdate segmentMetadataUpdate) + throws RemoteStorageException { + Objects.requireNonNull(segmentMetadataUpdate, "segmentMetadataUpdate can not be null"); + + lock.readLock().lock(); + try { + ensureInitializedAndNotClosed(); + + // Callers should use addRemoteLogSegmentMetadata to add RemoteLogSegmentMetadata with state as + // RemoteLogSegmentState.COPY_SEGMENT_STARTED. + if (segmentMetadataUpdate.state() == RemoteLogSegmentState.COPY_SEGMENT_STARTED) { + throw new IllegalArgumentException("Given remoteLogSegmentMetadata should not have the state as: " + + RemoteLogSegmentState.COPY_SEGMENT_STARTED); + } + + // Publish the message to the topic. + doPublishMetadata(segmentMetadataUpdate.remoteLogSegmentId().topicIdPartition(), segmentMetadataUpdate); + } finally { + lock.readLock().unlock(); + } + } + + @Override + public void putRemotePartitionDeleteMetadata(RemotePartitionDeleteMetadata remotePartitionDeleteMetadata) + throws RemoteStorageException { + Objects.requireNonNull(remotePartitionDeleteMetadata, "remotePartitionDeleteMetadata can not be null"); + + lock.readLock().lock(); + try { + ensureInitializedAndNotClosed(); + + doPublishMetadata(remotePartitionDeleteMetadata.topicIdPartition(), remotePartitionDeleteMetadata); + } finally { + lock.readLock().unlock(); + } + } + + private void doPublishMetadata(TopicIdPartition topicIdPartition, RemoteLogMetadata remoteLogMetadata) + throws RemoteStorageException { + log.debug("Publishing metadata for partition: [{}] with context: [{}]", topicIdPartition, remoteLogMetadata); + + try { + // Publish the message to the topic. + RecordMetadata recordMetadata = producerManager.publishMessage(remoteLogMetadata); + // Wait until the consumer catches up with this offset. This will ensure read-after-write consistency + // semantics. + consumerManager.waitTillConsumptionCatchesUp(recordMetadata); + } catch (KafkaException e) { + if (e instanceof RetriableException) { + throw e; + } else { + throw new RemoteStorageException(e); + } + } + } + + @Override + public Optional<RemoteLogSegmentMetadata> remoteLogSegmentMetadata(TopicIdPartition topicIdPartition, + int epochForOffset, + long offset) + throws RemoteStorageException { + lock.readLock().lock(); + try { + ensureInitializedAndNotClosed(); + + return remotePartitionMetadataStore.remoteLogSegmentMetadata(topicIdPartition, offset, epochForOffset); + } finally { + lock.readLock().unlock(); + } + } + + @Override + public Optional<Long> highestOffsetForEpoch(TopicIdPartition topicIdPartition, + int leaderEpoch) + throws RemoteStorageException { + lock.readLock().lock(); + try { + + ensureInitializedAndNotClosed(); + + return remotePartitionMetadataStore.highestLogOffset(topicIdPartition, leaderEpoch); + } finally { + lock.readLock().unlock(); + } + + } + + @Override + public Iterator<RemoteLogSegmentMetadata> listRemoteLogSegments(TopicIdPartition topicIdPartition) + throws RemoteStorageException { + Objects.requireNonNull(topicIdPartition, "topicIdPartition can not be null"); + + lock.readLock().lock(); + try { + ensureInitializedAndNotClosed(); + + return remotePartitionMetadataStore.listRemoteLogSegments(topicIdPartition); + } finally { + lock.readLock().unlock(); + } + } + + @Override + public Iterator<RemoteLogSegmentMetadata> listRemoteLogSegments(TopicIdPartition topicIdPartition, int leaderEpoch) + throws RemoteStorageException { + Objects.requireNonNull(topicIdPartition, "topicIdPartition can not be null"); + + lock.readLock().lock(); + try { + ensureInitializedAndNotClosed(); + + return remotePartitionMetadataStore.listRemoteLogSegments(topicIdPartition, leaderEpoch); + } finally { + lock.readLock().unlock(); + } + } + + public int metadataPartition(TopicIdPartition topicIdPartition) { + return rlmmTopicPartitioner.metadataPartition(topicIdPartition); + } + + // Visible For Testing + public Optional<Long> receivedOffsetForPartition(int metadataPartition) { + return consumerManager.receivedOffsetForPartition(metadataPartition); + } + + @Override + public void onPartitionLeadershipChanges(Set<TopicIdPartition> leaderPartitions, + Set<TopicIdPartition> followerPartitions) { + Objects.requireNonNull(leaderPartitions, "leaderPartitions can not be null"); + Objects.requireNonNull(followerPartitions, "followerPartitions can not be null"); + + log.info("Received leadership notifications with leader partitions {} and follower partitions {}", + leaderPartitions, followerPartitions); + + HashSet<TopicIdPartition> allPartitions = new HashSet<>(leaderPartitions); + allPartitions.addAll(followerPartitions); + lock.readLock().lock(); + try { + if (closing.get()) { + throw new IllegalStateException("This instance is in closing state"); + } + + if (!initialized.get()) { + // If it is not yet initialized, then keep them as pending partitions and assign them + // when it is initialized successfully in initializeResources(). + this.pendingAssignPartitions.addAll(allPartitions); + } else { + this.pendingAssignPartitions.clear(); + consumerManager.addAssignmentsForPartitions(allPartitions); + } + } finally { + lock.readLock().unlock(); + } + } + + @Override + public void onStopPartitions(Set<TopicIdPartition> partitions) { + lock.readLock().lock(); + try { + if (closing.get()) { + throw new IllegalStateException("This instance is in closing state"); + } + + if (!initialized.get()) { + // If it is not yet initialized, then remove them from the pending partitions if any. + if (!pendingAssignPartitions.isEmpty()) { + pendingAssignPartitions.removeAll(partitions); + } + } else { + consumerManager.removeAssignmentsForPartitions(partitions); + } + } finally { + lock.readLock().unlock(); + } + } + + @Override + public void configure(Map<String, ?> configs) { + Objects.requireNonNull(configs, "configs can not be null."); + + lock.writeLock().lock(); + try { + if (configured) { + log.info("Skipping configure as it is already configured."); + return; + } + + log.info("Started initializing with configs: {}", configs); + + rlmmConfig = new TopicBasedRemoteLogMetadataManagerConfig(configs); + rlmmTopicPartitioner = new RemoteLogMetadataTopicPartitioner(rlmmConfig.metadataTopicPartitionsCount()); + configured = true; + log.info("Successfully initialized with rlmmConfig: {}", rlmmConfig); + + // Scheduling the initialization producer/consumer managers in a separate thread. Required resources may + // not yet be available now. This thread makes sure that it is retried at regular intervals until it is + // successful. + initializationThread = KafkaThread.nonDaemon("RLMMInitializationThread", () -> initializeResources()); Review comment: @satishd : My understanding is that when we enable remote storage, we will do that through a rolling upgrade. So, at any given time, there is at most a single broker being down. Therefore, as long as the bootstrap broker list contains more than one broker, operations like creating topics can still be done while a single broker is starting. -- 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. To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org