nizhikov commented on code in PR #13247: URL: https://github.com/apache/kafka/pull/13247#discussion_r1347348232
########## tools/src/main/java/org/apache/kafka/tools/reassign/ReassignPartitionsCommand.java: ########## @@ -0,0 +1,1513 @@ +/* + * 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.tools.reassign; + +import com.fasterxml.jackson.core.JsonParseException; +import com.fasterxml.jackson.core.JsonProcessingException; +import com.fasterxml.jackson.databind.JsonMappingException; +import joptsimple.OptionSpec; +import org.apache.kafka.admin.AdminUtils; +import org.apache.kafka.admin.BrokerMetadata; +import org.apache.kafka.clients.admin.Admin; +import org.apache.kafka.clients.admin.AdminClientConfig; +import org.apache.kafka.clients.admin.AlterConfigOp; +import org.apache.kafka.clients.admin.ConfigEntry; +import org.apache.kafka.clients.admin.DescribeReplicaLogDirsResult; +import org.apache.kafka.clients.admin.NewPartitionReassignment; +import org.apache.kafka.clients.admin.PartitionReassignment; +import org.apache.kafka.clients.admin.TopicDescription; +import org.apache.kafka.common.KafkaFuture; +import org.apache.kafka.common.Node; +import org.apache.kafka.common.TopicPartition; +import org.apache.kafka.common.TopicPartitionReplica; +import org.apache.kafka.common.config.ConfigResource; +import org.apache.kafka.common.errors.ReplicaNotAvailableException; +import org.apache.kafka.common.errors.UnknownTopicOrPartitionException; +import org.apache.kafka.common.utils.Exit; +import org.apache.kafka.common.utils.Time; +import org.apache.kafka.common.utils.Utils; +import org.apache.kafka.server.common.AdminCommandFailedException; +import org.apache.kafka.server.common.AdminOperationException; +import org.apache.kafka.server.util.CommandLineUtils; +import org.apache.kafka.server.util.Json; +import org.apache.kafka.server.util.json.DecodeJson; +import org.apache.kafka.server.util.json.JsonObject; +import org.apache.kafka.server.util.json.JsonValue; +import org.apache.kafka.tools.TerseException; +import org.apache.kafka.tools.ToolsUtils; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collection; +import java.util.Collections; +import java.util.HashMap; +import java.util.HashSet; +import java.util.Iterator; +import java.util.LinkedHashMap; +import java.util.List; +import java.util.Map; +import java.util.Objects; +import java.util.Optional; +import java.util.Properties; +import java.util.Set; +import java.util.TreeSet; +import java.util.concurrent.ExecutionException; +import java.util.stream.Collectors; + +@SuppressWarnings("ClassDataAbstractionCoupling") +public class ReassignPartitionsCommand { + private static final String ANY_LOG_DIR = "any"; + + static final String HELP_TEXT = "This tool helps to move topic partitions between replicas."; + + private static final DecodeJson.DecodeInteger INT = new DecodeJson.DecodeInteger(); + + private static final DecodeJson.DecodeString STRING = new DecodeJson.DecodeString(); + + private static final DecodeJson<List<Integer>> INT_LIST = DecodeJson.decodeList(INT); + + private static final DecodeJson<List<String>> STRING_LIST = DecodeJson.decodeList(STRING); + + /** + * The earliest version of the partition reassignment JSON. We will default to this + * version if no other version number is given. + */ + static final int EARLIEST_VERSION = 1; + + /** + * The earliest version of the JSON for each partition reassignment topic. We will + * default to this version if no other version number is given. + */ + static final int EARLIEST_TOPICS_JSON_VERSION = 1; + + // Throttles that are set at the level of an individual broker. + //DynamicConfig.Broker.LeaderReplicationThrottledRateProp + static final String BROKER_LEVEL_LEADER_THROTTLE = "leader.replication.throttled.rate"; + //DynamicConfig.Broker.FollowerReplicationThrottledRateProp + static final String BROKER_LEVEL_FOLLOWER_THROTTLE = "follower.replication.throttled.rate"; + //DynamicConfig.Broker.ReplicaAlterLogDirsIoMaxBytesPerSecondProp + static final String BROKER_LEVEL_LOG_DIR_THROTTLE = "replica.alter.log.dirs.io.max.bytes.per.second"; + static final List<String> BROKER_LEVEL_THROTTLES = Arrays.asList( + BROKER_LEVEL_LEADER_THROTTLE, + BROKER_LEVEL_FOLLOWER_THROTTLE, + BROKER_LEVEL_LOG_DIR_THROTTLE + ); + + // Throttles that are set at the level of an individual topic. + //LogConfig.LEADER_REPLICATION_THROTTLED_REPLICAS_CONFIG + static final String TOPIC_LEVEL_LEADER_THROTTLE = "leader.replication.throttled.replicas"; + //LogConfig.FOLLOWER_REPLICATION_THROTTLED_REPLICAS_CONFIG + static final String TOPIC_LEVEL_FOLLOWER_THROTTLE = "follower.replication.throttled.replicas"; + private static final List<String> TOPIC_LEVEL_THROTTLES = Arrays.asList( + TOPIC_LEVEL_LEADER_THROTTLE, + TOPIC_LEVEL_FOLLOWER_THROTTLE + ); + + private static final String CANNOT_EXECUTE_BECAUSE_OF_EXISTING_MESSAGE = "Cannot execute because " + + "there is an existing partition assignment. Use --additional to override this and " + + "create a new partition assignment in addition to the existing one. The --additional " + + "flag can also be used to change the throttle by resubmitting the current reassignment."; + + private static final String YOU_MUST_RUN_VERIFY_PERIODICALLY_MESSAGE = "Warning: You must run " + + "--verify periodically, until the reassignment completes, to ensure the throttle " + + "is removed."; + + public static void main(String[] args) { + ReassignPartitionsCommandOptions opts = validateAndParseArgs(args); + boolean failed = true; + Admin adminClient = null; + + try { + Properties props = opts.options.has(opts.commandConfigOpt) + ? Utils.loadProps(opts.options.valueOf(opts.commandConfigOpt)) + : new Properties(); + props.put(AdminClientConfig.BOOTSTRAP_SERVERS_CONFIG, opts.options.valueOf(opts.bootstrapServerOpt)); + props.putIfAbsent(AdminClientConfig.CLIENT_ID_CONFIG, "reassign-partitions-tool"); + adminClient = Admin.create(props); + handleAction(adminClient, opts); + failed = false; + } catch (TerseException e) { + System.out.println(e.getMessage()); + } catch (Throwable e) { + System.out.println("Error: " + e.getMessage()); + System.out.println(Utils.stackTrace(e)); + } finally { + // It's good to do this after printing any error stack trace. + if (adminClient != null) { + adminClient.close(); + } + } + // If the command failed, exit with a non-zero exit code. + if (failed) { + Exit.exit(1); + } + } + + private static void handleAction(Admin adminClient, ReassignPartitionsCommandOptions opts) throws IOException, ExecutionException, InterruptedException, TerseException { + if (opts.options.has(opts.verifyOpt)) { + verifyAssignment(adminClient, + Utils.readFileAsString(opts.options.valueOf(opts.reassignmentJsonFileOpt)), + opts.options.has(opts.preserveThrottlesOpt)); + } else if (opts.options.has(opts.generateOpt)) { + generateAssignment(adminClient, + Utils.readFileAsString(opts.options.valueOf(opts.topicsToMoveJsonFileOpt)), + opts.options.valueOf(opts.brokerListOpt), + !opts.options.has(opts.disableRackAware)); + } else if (opts.options.has(opts.executeOpt)) { + executeAssignment(adminClient, + opts.options.has(opts.additionalOpt), + Utils.readFileAsString(opts.options.valueOf(opts.reassignmentJsonFileOpt)), + opts.options.valueOf(opts.interBrokerThrottleOpt), + opts.options.valueOf(opts.replicaAlterLogDirsThrottleOpt), + opts.options.valueOf(opts.timeoutOpt), + Time.SYSTEM); + } else if (opts.options.has(opts.cancelOpt)) { + cancelAssignment(adminClient, + Utils.readFileAsString(opts.options.valueOf(opts.reassignmentJsonFileOpt)), + opts.options.has(opts.preserveThrottlesOpt), + opts.options.valueOf(opts.timeoutOpt), + Time.SYSTEM); + } else if (opts.options.has(opts.listOpt)) { + listReassignments(adminClient); + } else { + throw new RuntimeException("Unsupported action."); + } + } + + /** + * The entry point for the --verify command. + * + * @param adminClient The AdminClient to use. + * @param jsonString The JSON string to use for the topics and partitions to verify. + * @param preserveThrottles True if we should avoid changing topic or broker throttles. + * + * @return A result that is useful for testing. + */ + static VerifyAssignmentResult verifyAssignment(Admin adminClient, + String jsonString, + Boolean preserveThrottles + ) throws ExecutionException, InterruptedException, JsonProcessingException { + Tuple2<List<Tuple2<TopicPartition, List<Integer>>>, Map<TopicPartitionReplica, String>> t0 = parsePartitionReassignmentData(jsonString); + + List<Tuple2<TopicPartition, List<Integer>>> targetParts = t0.v1; + Map<TopicPartitionReplica, String> targetLogDirs = t0.v2; + + Tuple2<Map<TopicPartition, PartitionReassignmentState>, Boolean> t1 = verifyPartitionAssignments(adminClient, targetParts); + + Map<TopicPartition, PartitionReassignmentState> partStates = t1.v1; + Boolean partsOngoing = t1.v2; + + Tuple2<Map<TopicPartitionReplica, LogDirMoveState>, Boolean> t2 = verifyReplicaMoves(adminClient, targetLogDirs); + + Map<TopicPartitionReplica, LogDirMoveState> moveStates = t2.v1; + Boolean movesOngoing = t2.v2; + + if (!partsOngoing && !movesOngoing && !preserveThrottles) { + // If the partition assignments and replica assignments are done, clear any throttles + // that were set. We have to clear all throttles, because we don't have enough + // information to know all of the source brokers that might have been involved in the + // previous reassignments. + clearAllThrottles(adminClient, targetParts); + } + + return new VerifyAssignmentResult(partStates, partsOngoing, moveStates, movesOngoing); + } + + /** + * Verify the partition reassignments specified by the user. + * + * @param adminClient The AdminClient to use. + * @param targets The partition reassignments specified by the user. + * + * @return A tuple of the partition reassignment states, and a + * boolean which is true if there are no ongoing + * reassignments (including reassignments not described + * in the JSON file.) + */ + private static Tuple2<Map<TopicPartition, PartitionReassignmentState>, Boolean> verifyPartitionAssignments(Admin adminClient, + List<Tuple2<TopicPartition, List<Integer>>> targets + ) throws ExecutionException, InterruptedException { + Tuple2<Map<TopicPartition, PartitionReassignmentState>, Boolean> t0 = findPartitionReassignmentStates(adminClient, targets); + System.out.println(partitionReassignmentStatesToString(t0.v1)); + return t0; + } + + static int compareTopicPartitions(TopicPartition a, TopicPartition b) { + int topicOrder = Objects.compare(a.topic(), b.topic(), String::compareTo); + return topicOrder == 0 ? Integer.compare(a.partition(), b.partition()) : topicOrder; + } + + static int compareTopicPartitionReplicas(TopicPartitionReplica a, TopicPartitionReplica b) { + int brokerOrder = Integer.compare(a.brokerId(), b.brokerId()); + + if (brokerOrder != 0) + return brokerOrder; + + int topicOrder = Objects.compare(a.topic(), b.topic(), String::compareTo); + return topicOrder == 0 ? Integer.compare(a.partition(), b.partition()) : topicOrder; + } + + /** + * Convert partition reassignment states to a human-readable string. + * + * @param states A map from topic partitions to states. + * @return A string summarizing the partition reassignment states. + */ + static String partitionReassignmentStatesToString(Map<TopicPartition, PartitionReassignmentState> states) { + List<String> bld = new ArrayList<>(); + bld.add("Status of partition reassignment:"); + states.keySet().stream().sorted(ReassignPartitionsCommand::compareTopicPartitions).forEach(topicPartition -> { + PartitionReassignmentState state = states.get(topicPartition); + if (state.done) { + if (state.currentReplicas.equals(state.targetReplicas)) { + bld.add(String.format("Reassignment of partition %s is completed.", topicPartition)); + } else { + String currentReplicaStr = state.currentReplicas.stream().map(String::valueOf).collect(Collectors.joining(",")); + String targetReplicaStr = state.targetReplicas.stream().map(String::valueOf).collect(Collectors.joining(",")); + + bld.add("There is no active reassignment of partition " + topicPartition + ", " + + "but replica set is " + currentReplicaStr + " rather than " + + targetReplicaStr + "."); + } + } else { + bld.add(String.format("Reassignment of partition %s is still in progress.", topicPartition)); + } + }); + return bld.stream().collect(Collectors.joining(System.lineSeparator())); + } + + /** + * Find the state of the specified partition reassignments. + * + * @param adminClient The Admin client to use. + * @param targetReassignments The reassignments we want to learn about. + * + * @return A tuple containing the reassignment states for each topic + * partition, plus whether there are any ongoing reassignments. + */ + static Tuple2<Map<TopicPartition, PartitionReassignmentState>, Boolean> findPartitionReassignmentStates(Admin adminClient, + List<Tuple2<TopicPartition, + List<Integer>>> targetReassignments + ) throws ExecutionException, InterruptedException { + Map<TopicPartition, PartitionReassignment> currentReassignments = adminClient. + listPartitionReassignments().reassignments().get(); + + List<Tuple2<TopicPartition, List<Integer>>> foundReassignments = new ArrayList<>(); + List<Tuple2<TopicPartition, List<Integer>>> notFoundReassignments = new ArrayList<>(); + + targetReassignments.forEach(reassignment -> { + if (currentReassignments.containsKey(reassignment.v1)) + foundReassignments.add(reassignment); + else + notFoundReassignments.add(reassignment); + }); + + List<Tuple2<TopicPartition, PartitionReassignmentState>> foundResults = foundReassignments.stream().map(e -> { + TopicPartition part = e.v1; + List<Integer> targetReplicas = e.v2; + return new Tuple2<>(part, + new PartitionReassignmentState( + currentReassignments.get(part).replicas(), + targetReplicas, + false)); + }).collect(Collectors.toList()); + + Set<String> topicNamesToLookUp = notFoundReassignments.stream() + .map(e -> e.v1) + .filter(part -> !currentReassignments.containsKey(part)) + .map(TopicPartition::topic) + .collect(Collectors.toSet()); + + Map<String, KafkaFuture<TopicDescription>> topicDescriptions = adminClient. + describeTopics(topicNamesToLookUp).topicNameValues(); + + List<Tuple2<TopicPartition, PartitionReassignmentState>> notFoundResults = new ArrayList<>(); + for (Tuple2<TopicPartition, List<Integer>> e : notFoundReassignments) { + TopicPartition part = e.v1; + List<Integer> targetReplicas = e.v2; + + if (currentReassignments.containsKey(part)) { + PartitionReassignment reassignment = currentReassignments.get(part); + notFoundResults.add(new Tuple2<>(part, new PartitionReassignmentState( + reassignment.replicas(), + targetReplicas, + false))); + } else { + notFoundResults.add(new Tuple2<>(part, topicDescriptionFutureToState(part.partition(), + topicDescriptions.get(part.topic()), targetReplicas))); + } + } + + Map<TopicPartition, PartitionReassignmentState> allResults = new HashMap<>(); + foundResults.forEach(e -> allResults.put(e.v1, e.v2)); + notFoundResults.forEach(e -> allResults.put(e.v1, e.v2)); + + return new Tuple2<>(allResults, currentReassignments.size() > 0); + } + + private static PartitionReassignmentState topicDescriptionFutureToState(int partition, + KafkaFuture<TopicDescription> future, + List<Integer> targetReplicas + ) throws InterruptedException, ExecutionException { + try { + TopicDescription topicDescription = future.get(); + if (topicDescription.partitions().size() < partition) { + throw new ExecutionException("Too few partitions found", new UnknownTopicOrPartitionException()); + } + return new PartitionReassignmentState( + topicDescription.partitions().get(partition).replicas().stream().map(Node::id).collect(Collectors.toList()), + targetReplicas, + true); + } catch (ExecutionException t) { + if (t.getCause() instanceof UnknownTopicOrPartitionException) + return new PartitionReassignmentState(Collections.emptyList(), targetReplicas, true); + + throw t; + } + } + + /** + * Verify the replica reassignments specified by the user. + * + * @param adminClient The AdminClient to use. + * @param targetReassignments The replica reassignments specified by the user. + * + * @return A tuple of the replica states, and a boolean which is true + * if there are any ongoing replica moves. + * + * Note: Unlike in verifyPartitionAssignments, we will + * return false here even if there are unrelated ongoing + * reassignments. (We don't have an efficient API that + * returns all ongoing replica reassignments.) + */ + private static Tuple2<Map<TopicPartitionReplica, LogDirMoveState>, Boolean> verifyReplicaMoves(Admin adminClient, + Map<TopicPartitionReplica, String> targetReassignments + ) throws ExecutionException, InterruptedException { + Map<TopicPartitionReplica, LogDirMoveState> moveStates = findLogDirMoveStates(adminClient, targetReassignments); + System.out.println(replicaMoveStatesToString(moveStates)); + return new Tuple2<>(moveStates, !moveStates.values().stream().allMatch(LogDirMoveState::done)); + } + + /** + * Find the state of the specified partition reassignments. + * + * @param adminClient The AdminClient to use. + * @param targetMoves The movements we want to learn about. The map is keyed + * by TopicPartitionReplica, and its values are target log + * directories. + * + * @return The states for each replica movement. + */ + static Map<TopicPartitionReplica, LogDirMoveState> findLogDirMoveStates(Admin adminClient, + Map<TopicPartitionReplica, String> targetMoves + ) throws ExecutionException, InterruptedException { + Map<TopicPartitionReplica, DescribeReplicaLogDirsResult.ReplicaLogDirInfo> replicaLogDirInfos = adminClient + .describeReplicaLogDirs(targetMoves.keySet()).all().get(); + + return targetMoves.entrySet().stream().collect(Collectors.toMap(Map.Entry::getKey, e -> { + TopicPartitionReplica replica = e.getKey(); + String targetLogDir = e.getValue(); + + if (!replicaLogDirInfos.containsKey(replica)) + return new MissingReplicaMoveState(targetLogDir); + + DescribeReplicaLogDirsResult.ReplicaLogDirInfo info = replicaLogDirInfos.get(replica); + + if (info.getCurrentReplicaLogDir() == null) + return new MissingLogDirMoveState(targetLogDir); + + if (info.getFutureReplicaLogDir() == null) { + if (info.getCurrentReplicaLogDir().equals(targetLogDir)) + return new CompletedMoveState(targetLogDir); + + return new CancelledMoveState(info.getCurrentReplicaLogDir(), targetLogDir); + } + + return new ActiveMoveState(info.getCurrentReplicaLogDir(), targetLogDir, info.getFutureReplicaLogDir()); + })); + } + + /** + * Convert replica move states to a human-readable string. + * + * @param states A map from topic partition replicas to states. + * @return A tuple of a summary string, and a boolean describing + * whether there are any active replica moves. + */ + static String replicaMoveStatesToString(Map<TopicPartitionReplica, LogDirMoveState> states) { + List<String> bld = new ArrayList<>(); + states.keySet().stream().sorted(ReassignPartitionsCommand::compareTopicPartitionReplicas).forEach(replica -> { + LogDirMoveState state = states.get(replica); + if (state instanceof MissingLogDirMoveState) { + bld.add("Partition " + replica.topic() + "-" + replica.partition() + " is not found " + + "in any live log dir on broker " + replica.brokerId() + ". There is likely an " + + "offline log directory on the broker."); + } else if (state instanceof MissingReplicaMoveState) { + bld.add("Partition " + replica.topic() + "-" + replica.partition() + " cannot be found " + + "in any live log directory on broker " + replica.brokerId() + "."); + } else if (state instanceof ActiveMoveState) { + String targetLogDir = ((ActiveMoveState) state).targetLogDir; + String futureLogDir = ((ActiveMoveState) state).futureLogDir; + if (targetLogDir.equals(futureLogDir)) { + bld.add("Reassignment of replica " + replica + " is still in progress."); + } else { + bld.add("Partition " + replica.topic() + "-" + replica.partition() + " on broker " + + replica.brokerId() + " is being moved to log dir " + futureLogDir + " " + + "instead of " + targetLogDir + "."); + } + } else if (state instanceof CancelledMoveState) { + String targetLogDir = ((CancelledMoveState) state).targetLogDir; + String currentLogDir = ((CancelledMoveState) state).currentLogDir; + bld.add("Partition " + replica.topic() + "-" + replica.partition() + " on broker " + + replica.brokerId() + " is not being moved from log dir " + currentLogDir + " to " + + targetLogDir + "."); + } else if (state instanceof CompletedMoveState) { + bld.add("Reassignment of replica " + replica + " completed successfully."); + } + }); + + return bld.stream().collect(Collectors.joining(System.lineSeparator())); + } + + /** + * Clear all topic-level and broker-level throttles. + * + * @param adminClient The AdminClient to use. + * @param targetParts The target partitions loaded from the JSON file. + */ + private static void clearAllThrottles(Admin adminClient, + List<Tuple2<TopicPartition, List<Integer>>> targetParts + ) throws ExecutionException, InterruptedException { + Set<Integer> activeBrokers = adminClient.describeCluster().nodes().get().stream().map(Node::id).collect(Collectors.toSet()); + Set<Integer> brokers = new HashSet<>(activeBrokers); + targetParts.forEach(t -> brokers.addAll(t.v2)); + + System.out.printf("Clearing broker-level throttles on broker%s %s%n", + brokers.size() == 1 ? "" : "s", Utils.join(brokers, ",")); + clearBrokerLevelThrottles(adminClient, brokers); + + Set<String> topics = targetParts.stream().map(t -> t.v1.topic()).collect(Collectors.toSet()); + System.out.printf("Clearing topic-level throttles on topic%s %s%n", + topics.size() == 1 ? "" : "s", Utils.join(topics, ",")); + clearTopicLevelThrottles(adminClient, topics); + } + + /** + * Clear all throttles which have been set at the broker level. + * + * @param adminClient The AdminClient to use. + * @param brokers The brokers to clear the throttles for. + */ + private static void clearBrokerLevelThrottles(Admin adminClient, Set<Integer> brokers) throws ExecutionException, InterruptedException { + Map<ConfigResource, Collection<AlterConfigOp>> configOps = new HashMap<>(); + brokers.forEach(brokerId -> configOps.put( + new ConfigResource(ConfigResource.Type.BROKER, brokerId.toString()), + BROKER_LEVEL_THROTTLES.stream().map(throttle -> new AlterConfigOp( + new ConfigEntry(throttle, null), AlterConfigOp.OpType.DELETE)).collect(Collectors.toList()) + )); + adminClient.incrementalAlterConfigs(configOps).all().get(); + } + + /** + * Clear the reassignment throttles for the specified topics. + * + * @param adminClient The AdminClient to use. + * @param topics The topics to clear the throttles for. + */ + private static void clearTopicLevelThrottles(Admin adminClient, Set<String> topics) throws ExecutionException, InterruptedException { + Map<ConfigResource, Collection<AlterConfigOp>> configOps = topics.stream().collect(Collectors.toMap( + topicName -> new ConfigResource(ConfigResource.Type.TOPIC, topicName), + topicName -> TOPIC_LEVEL_THROTTLES.stream().map(throttle -> new AlterConfigOp(new ConfigEntry(throttle, null), + AlterConfigOp.OpType.DELETE)).collect(Collectors.toList()) + )); + + adminClient.incrementalAlterConfigs(configOps).all().get(); + } + + /** + * The entry point for the --generate command. + * + * @param adminClient The AdminClient to use. + * @param reassignmentJson The JSON string to use for the topics to reassign. + * @param brokerListString The comma-separated string of broker IDs to use. + * @param enableRackAwareness True if rack-awareness should be enabled. + * + * @return A tuple containing the proposed assignment and the + * current assignment. + */ + static Tuple2<Map<TopicPartition, List<Integer>>, Map<TopicPartition, List<Integer>>> generateAssignment(Admin adminClient, + String reassignmentJson, + String brokerListString, + Boolean enableRackAwareness + ) throws ExecutionException, InterruptedException, JsonProcessingException { + Tuple2<List<Integer>, List<String>> t0 = parseGenerateAssignmentArgs(reassignmentJson, brokerListString); + + List<Integer> brokersToReassign = t0.v1; + List<String> topicsToReassign = t0.v2; + + Map<TopicPartition, List<Integer>> currentAssignments = getReplicaAssignmentForTopics(adminClient, topicsToReassign); + List<BrokerMetadata> brokerMetadatas = getBrokerMetadata(adminClient, brokersToReassign, enableRackAwareness); + Map<TopicPartition, List<Integer>> proposedAssignments = calculateAssignment(currentAssignments, brokerMetadatas); + System.out.printf("Current partition replica assignment%n%s%n%n", + formatAsReassignmentJson(currentAssignments, Collections.emptyMap())); + System.out.printf("Proposed partition reassignment configuration%n%s%n", + formatAsReassignmentJson(proposedAssignments, Collections.emptyMap())); + return new Tuple2<>(proposedAssignments, currentAssignments); + } + + /** + * Calculate the new partition assignments to suggest in --generate. + * + * @param currentAssignment The current partition assignments. + * @param brokerMetadatas The rack information for each broker. + * + * @return A map from partitions to the proposed assignments for each. + */ + private static Map<TopicPartition, List<Integer>> calculateAssignment(Map<TopicPartition, List<Integer>> currentAssignment, + List<BrokerMetadata> brokerMetadatas) { + Map<String, List<Map.Entry<TopicPartition, List<Integer>>>> groupedByTopic = new HashMap<>(); + for (Map.Entry<TopicPartition, List<Integer>> e : currentAssignment.entrySet()) + groupedByTopic.computeIfAbsent(e.getKey().topic(), k -> new ArrayList<>()).add(e); + Map<TopicPartition, List<Integer>> proposedAssignments = new HashMap<>(); + groupedByTopic.forEach((topic, assignment) -> { + List<Integer> replicas = assignment.get(0).getValue(); + Map<Integer, List<Integer>> assignedReplicas = AdminUtils. + assignReplicasToBrokers(brokerMetadatas, assignment.size(), replicas.size()); + assignedReplicas.forEach((partition, replicas0) -> + proposedAssignments.put(new TopicPartition(topic, partition), replicas0)); + }); + return proposedAssignments; + } + + private static Map<String, TopicDescription> describeTopics(Admin adminClient, + Set<String> topics) throws ExecutionException, InterruptedException { + Map<String, KafkaFuture<TopicDescription>> futures = adminClient.describeTopics(topics).topicNameValues(); + Map<String, TopicDescription> res = new HashMap<>(); + for (Map.Entry<String, KafkaFuture<TopicDescription>> e : futures.entrySet()) { + String topicName = e.getKey(); + KafkaFuture<TopicDescription> topicDescriptionFuture = e.getValue(); + try { + res.put(topicName, topicDescriptionFuture.get()); + } catch (ExecutionException t) { + if (t.getCause() instanceof UnknownTopicOrPartitionException) + throw new ExecutionException( + new UnknownTopicOrPartitionException("Topic " + topicName + " not found.")); + throw t; + } + } + return res; + } + + /** + * Get the current replica assignments for some topics. + * + * @param adminClient The AdminClient to use. + * @param topics The topics to get information about. + * @return A map from partitions to broker assignments. + * If any topic can't be found, an exception will be thrown. + */ + static Map<TopicPartition, List<Integer>> getReplicaAssignmentForTopics(Admin adminClient, + List<String> topics + ) throws ExecutionException, InterruptedException { + Map<TopicPartition, List<Integer>> res = new HashMap<>(); + describeTopics(adminClient, new HashSet<>(topics)).forEach((topicName, topicDescription) -> + topicDescription.partitions().forEach(info -> res.put( + new TopicPartition(topicName, info.partition()), + info.replicas().stream().map(Node::id).collect(Collectors.toList()) + ) + )); + return res; + } + + /** + * Get the current replica assignments for some partitions. + * + * @param adminClient The AdminClient to use. + * @param partitions The partitions to get information about. + * @return A map from partitions to broker assignments. + * If any topic can't be found, an exception will be thrown. + */ + static Map<TopicPartition, List<Integer>> getReplicaAssignmentForPartitions(Admin adminClient, + Set<TopicPartition> partitions + ) throws ExecutionException, InterruptedException { + Map<TopicPartition, List<Integer>> res = new HashMap<>(); + describeTopics(adminClient, partitions.stream().map(TopicPartition::topic).collect(Collectors.toSet())).forEach((topicName, topicDescription) -> { + topicDescription.partitions().forEach(info -> { + TopicPartition tp = new TopicPartition(topicName, info.partition()); + if (partitions.contains(tp)) + res.put(tp, info.replicas().stream().map(Node::id).collect(Collectors.toList())); + }); + }); + return res; + } + + /** + * Find the rack information for some brokers. + * + * @param adminClient The AdminClient object. + * @param brokers The brokers to gather metadata about. + * @param enableRackAwareness True if we should return rack information, and throw an + * exception if it is inconsistent. + * + * @return The metadata for each broker that was found. + * Brokers that were not found will be omitted. + */ + static List<BrokerMetadata> getBrokerMetadata(Admin adminClient, List<Integer> brokers, boolean enableRackAwareness) throws ExecutionException, InterruptedException { + Set<Integer> brokerSet = new HashSet<>(brokers); + List<BrokerMetadata> results = adminClient.describeCluster().nodes().get().stream() + .filter(node -> brokerSet.contains(node.id())) + .map(node -> (enableRackAwareness && node.rack() != null) + ? new BrokerMetadata(node.id(), Optional.of(node.rack())) + : new BrokerMetadata(node.id(), Optional.empty()) + ).collect(Collectors.toList()); + + long numRackless = results.stream().filter(m -> !m.rack.isPresent()).count(); + if (enableRackAwareness && numRackless != 0 && numRackless != results.size()) { + throw new AdminOperationException("Not all brokers have rack information. Add " + + "--disable-rack-aware in command line to make replica assignment without rack " + + "information."); + } + return results; + } + + /** + * Parse and validate data gathered from the command-line for --generate + * In particular, we parse the JSON and validate that duplicate brokers and + * topics don't appear. + * + * @param reassignmentJson The JSON passed to --generate . + * @param brokerList A list of brokers passed to --generate. + * + * @return A tuple of brokers to reassign, topics to reassign + */ + static Tuple2<List<Integer>, List<String>> parseGenerateAssignmentArgs(String reassignmentJson, + String brokerList) throws JsonMappingException { + List<Integer> brokerListToReassign = Arrays.stream(brokerList.split(",")).map(Integer::parseInt).collect(Collectors.toList()); + Set<Integer> duplicateReassignments = ToolsUtils.duplicates(brokerListToReassign); + if (!duplicateReassignments.isEmpty()) + throw new AdminCommandFailedException(String.format("Broker list contains duplicate entries: %s", duplicateReassignments)); + List<String> topicsToReassign = parseTopicsData(reassignmentJson); + Set<String> duplicateTopicsToReassign = ToolsUtils.duplicates(topicsToReassign); + if (!duplicateTopicsToReassign.isEmpty()) + throw new AdminCommandFailedException(String.format("List of topics to reassign contains duplicate entries: %s", + duplicateTopicsToReassign)); + return new Tuple2<>(brokerListToReassign, topicsToReassign); + } + + /** + * The entry point for the --execute and --execute-additional commands. + * + * @param adminClient The AdminClient to use. + * @param additional Whether --additional was passed. + * @param reassignmentJson The JSON string to use for the topics to reassign. + * @param interBrokerThrottle The inter-broker throttle to use, or a negative + * number to skip using a throttle. + * @param logDirThrottle The replica log directory throttle to use, or a + * negative number to skip using a throttle. + * @param timeoutMs The maximum time in ms to wait for log directory + * replica assignment to begin. + * @param time The Time object to use. + */ + static void executeAssignment(Admin adminClient, + Boolean additional, + String reassignmentJson, + Long interBrokerThrottle, + Long logDirThrottle, + Long timeoutMs, + Time time + ) throws ExecutionException, InterruptedException, JsonProcessingException, TerseException { + Tuple2<Map<TopicPartition, List<Integer>>, Map<TopicPartitionReplica, String>> t0 = parseExecuteAssignmentArgs(reassignmentJson); + + Map<TopicPartition, List<Integer>> proposedParts = t0.v1; + Map<TopicPartitionReplica, String> proposedReplicas = t0.v2; + Map<TopicPartition, PartitionReassignment> currentReassignments = adminClient. + listPartitionReassignments().reassignments().get(); + // If there is an existing assignment, check for --additional before proceeding. + // This helps avoid surprising users. + if (!additional && !currentReassignments.isEmpty()) { + throw new TerseException(CANNOT_EXECUTE_BECAUSE_OF_EXISTING_MESSAGE); + } + Set<Integer> brokers = new HashSet<>(); + proposedParts.values().forEach(brokers::addAll); + + verifyBrokerIds(adminClient, brokers); + Map<TopicPartition, List<Integer>> currentParts = getReplicaAssignmentForPartitions(adminClient, proposedParts.keySet()); + System.out.println(currentPartitionReplicaAssignmentToString(proposedParts, currentParts)); + + if (interBrokerThrottle >= 0 || logDirThrottle >= 0) { + System.out.println(YOU_MUST_RUN_VERIFY_PERIODICALLY_MESSAGE); + + if (interBrokerThrottle >= 0) { + Map<String, Map<Integer, PartitionMove>> moveMap = calculateProposedMoveMap(currentReassignments, proposedParts, currentParts); + modifyReassignmentThrottle(adminClient, moveMap, interBrokerThrottle); + } + + if (logDirThrottle >= 0) { + Set<Integer> movingBrokers = calculateMovingBrokers(proposedReplicas.keySet()); + modifyLogDirThrottle(adminClient, movingBrokers, logDirThrottle); + } + } + + // Execute the partition reassignments. + Map<TopicPartition, Throwable> errors = alterPartitionReassignments(adminClient, proposedParts); + if (!errors.isEmpty()) { + throw new TerseException( + String.format("Error reassigning partition(s):%n%s", + errors.keySet().stream() + .sorted(ReassignPartitionsCommand::compareTopicPartitions) + .map(part -> part + ": " + errors.get(part).getMessage()) + .collect(Collectors.joining(System.lineSeparator())))); + } + System.out.printf("Successfully started partition reassignment%s for %s%n", + proposedParts.size() == 1 ? "" : "s", + proposedParts.keySet().stream() + .sorted(ReassignPartitionsCommand::compareTopicPartitions) + .map(Objects::toString) + .collect(Collectors.joining(","))); + if (!proposedReplicas.isEmpty()) { + executeMoves(adminClient, proposedReplicas, timeoutMs, time); + } + } + + /** + * Execute some partition log directory movements. + * + * @param adminClient The AdminClient to use. + * @param proposedReplicas A map from TopicPartitionReplicas to the + * directories to move them to. + * @param timeoutMs The maximum time in ms to wait for log directory + * replica assignment to begin. + * @param time The Time object to use. + */ + private static void executeMoves(Admin adminClient, + Map<TopicPartitionReplica, String> proposedReplicas, + Long timeoutMs, + Time time + ) throws InterruptedException, TerseException { + long startTimeMs = time.milliseconds(); + Map<TopicPartitionReplica, String> pendingReplicas = new HashMap<>(); + pendingReplicas.putAll(proposedReplicas); + boolean done = false; + do { + Set<TopicPartitionReplica> completed = alterReplicaLogDirs(adminClient, pendingReplicas); + if (!completed.isEmpty()) { + System.out.printf("Successfully started log directory move%s for: %s%n", + completed.size() == 1 ? "" : "s", + completed.stream() + .sorted(ReassignPartitionsCommand::compareTopicPartitionReplicas) + .map(Object::toString) + .collect(Collectors.joining(","))); + } + completed.forEach(pendingReplicas::remove); + if (pendingReplicas.isEmpty()) { + done = true; + } else if (time.milliseconds() >= startTimeMs + timeoutMs) { + throw new TerseException(String.format( + "Timed out before log directory move%s could be started for: %s", + pendingReplicas.size() == 1 ? "" : "s", + pendingReplicas.keySet().stream() + .sorted(ReassignPartitionsCommand::compareTopicPartitionReplicas) + .map(Object::toString) + .collect(Collectors.joining(",")))); + } else { + // If a replica has been moved to a new host and we also specified a particular + // log directory, we will have to keep retrying the alterReplicaLogDirs + // call. It can't take effect until the replica is moved to that host. + time.sleep(100); + } + + + } while (!done); + } + + /** + * Entry point for the --list command. + * + * @param adminClient The AdminClient to use. + */ + private static void listReassignments(Admin adminClient) throws ExecutionException, InterruptedException { + System.out.println(curReassignmentsToString(adminClient)); + } + + /** + * Convert the current partition reassignments to text. + * + * @param adminClient The AdminClient to use. + * @return A string describing the current partition reassignments. + */ + static String curReassignmentsToString(Admin adminClient) throws ExecutionException, InterruptedException { + Map<TopicPartition, PartitionReassignment> currentReassignments = adminClient. + listPartitionReassignments().reassignments().get(); + String text = currentReassignments.keySet().stream().sorted(ReassignPartitionsCommand::compareTopicPartitions).map(part -> { + PartitionReassignment reassignment = currentReassignments.get(part); + List<Integer> replicas = reassignment.replicas(); + List<Integer> addingReplicas = reassignment.addingReplicas(); + List<Integer> removingReplicas = reassignment.removingReplicas(); + + return String.format("%s: replicas: %s.%s%s", + part, + replicas.stream().map(Object::toString).collect(Collectors.joining(",")), + addingReplicas.isEmpty() ? "" : String.format(" adding: %s.", addingReplicas.stream() + .map(Object::toString) + .collect(Collectors.joining(","))), + removingReplicas.isEmpty() ? "" : String.format(" removing: %s.", removingReplicas.stream() + .map(Object::toString) + .collect(Collectors.joining(","))) + ); + }).collect(Collectors.joining(System.lineSeparator())); + + return text.isEmpty() + ? "No partition reassignments found." + : String.format("Current partition reassignments:%n%s", text); + } + + /** + * Verify that all the brokers in an assignment exist. + * + * @param adminClient The AdminClient to use. + * @param brokers The broker IDs to verify. + */ + private static void verifyBrokerIds(Admin adminClient, Set<Integer> brokers) throws ExecutionException, InterruptedException { + Set<Integer> allNodeIds = adminClient.describeCluster().nodes().get().stream().map(Node::id).collect(Collectors.toSet()); + Optional<Integer> unknown = brokers.stream() + .filter(brokerId -> !allNodeIds.contains(brokerId)) + .findFirst(); + if (unknown.isPresent()) + throw new AdminCommandFailedException("Unknown broker id " + unknown.get()); + } + + /** + * Return the string which we want to print to describe the current partition assignment. + * + * @param proposedParts The proposed partition assignment. + * @param currentParts The current partition assignment. + * + * @return The string to print. We will only print information about + * partitions that appear in the proposed partition assignment. + */ + static String currentPartitionReplicaAssignmentToString(Map<TopicPartition, List<Integer>> proposedParts, + Map<TopicPartition, List<Integer>> currentParts) throws JsonProcessingException { + Map<TopicPartition, List<Integer>> partitionsToBeReassigned = currentParts.entrySet().stream() + .filter(e -> proposedParts.containsKey(e.getKey())) + .collect(Collectors.toMap(Map.Entry::getKey, Map.Entry::getValue)); + + return String.format("Current partition replica assignment%n%n%s%n%nSave this to use as the %s", + formatAsReassignmentJson(partitionsToBeReassigned, Collections.emptyMap()), + "--reassignment-json-file option during rollback"); + } + + /** + * Execute the given partition reassignments. + * + * @param adminClient The admin client object to use. + * @param reassignments A map from topic names to target replica assignments. + * @return A map from partition objects to error strings. + */ + static Map<TopicPartition, Throwable> alterPartitionReassignments(Admin adminClient, + Map<TopicPartition, List<Integer>> reassignments) throws InterruptedException { + Map<TopicPartition, Optional<NewPartitionReassignment>> args = new HashMap<>(); + reassignments.entrySet().forEach(e -> { + TopicPartition part = e.getKey(); + List<Integer> replicas = e.getValue(); + + args.put(part, Optional.of(new NewPartitionReassignment(replicas))); + }); + Map<TopicPartition, KafkaFuture<Void>> results = adminClient.alterPartitionReassignments(args).values(); + Map<TopicPartition, Throwable> errors = new HashMap<>(); + for (Map.Entry<TopicPartition, KafkaFuture<Void>> e : results.entrySet()) { + try { + e.getValue().get(); + } catch (ExecutionException t) { + errors.put(e.getKey(), t.getCause()); + } + } + return errors; + } + + /** + * Cancel the given partition reassignments. + * + * @param adminClient The admin client object to use. + * @param reassignments The partition reassignments to cancel. + * @return A map from partition objects to error strings. + */ + static Map<TopicPartition, Throwable> cancelPartitionReassignments(Admin adminClient, + Set<TopicPartition> reassignments) throws InterruptedException { + Map<TopicPartition, Optional<NewPartitionReassignment>> args = new HashMap<>(); + reassignments.forEach(part -> args.put(part, Optional.empty())); + + Map<TopicPartition, KafkaFuture<Void>> results = adminClient.alterPartitionReassignments(args).values(); + Map<TopicPartition, Throwable> errors = new HashMap<>(); + for (Map.Entry<TopicPartition, KafkaFuture<Void>> e : results.entrySet()) { + try { + e.getValue().get(); + } catch (ExecutionException t) { + errors.put(e.getKey(), t.getCause()); + } + } + return errors; + } + + /** + * Compute the in progress partition move from the current reassignments. + * @param currentReassignments All replicas, adding replicas and removing replicas of target partitions + */ + private static Map<String, Map<Integer, PartitionMove>> calculateCurrentMoveMap(Map<TopicPartition, PartitionReassignment> currentReassignments) { + Map<String, Map<Integer, PartitionMove>> moveMap = new HashMap<>(); + // Add the current reassignments to the move map. + currentReassignments.entrySet().forEach(e -> { + TopicPartition part = e.getKey(); + PartitionReassignment reassignment = e.getValue(); + + List<Integer> allReplicas = reassignment.replicas(); + List<Integer> addingReplicas = reassignment.addingReplicas(); + + // The addingReplicas is included in the replicas during reassignment + Set<Integer> sources = new HashSet<>(allReplicas); + sources.removeAll(addingReplicas); + + Set<Integer> destinations = new HashSet<>(addingReplicas); + + Map<Integer, PartitionMove> partMoves = moveMap.computeIfAbsent(part.topic(), k -> new HashMap<>()); + partMoves.put(part.partition(), new PartitionMove(sources, destinations)); + }); Review Comment: Fixed -- 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