cychiu8 commented on code in PR #22252: URL: https://github.com/apache/kafka/pull/22252#discussion_r3234897473
########## server/src/test/java/org/apache/kafka/server/requests/FetchRequestTest.java: ########## @@ -0,0 +1,736 @@ +/* + * 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.requests; + +import org.apache.kafka.clients.admin.Admin; +import org.apache.kafka.clients.admin.NewPartitionReassignment; +import org.apache.kafka.clients.consumer.ConsumerConfig; +import org.apache.kafka.clients.consumer.KafkaConsumer; +import org.apache.kafka.clients.consumer.RangeAssignor; +import org.apache.kafka.clients.producer.Producer; +import org.apache.kafka.clients.producer.ProducerRecord; +import org.apache.kafka.common.TopicIdPartition; +import org.apache.kafka.common.TopicPartition; +import org.apache.kafka.common.Uuid; +import org.apache.kafka.common.message.FetchRequestData; +import org.apache.kafka.common.protocol.ApiKeys; +import org.apache.kafka.common.protocol.Errors; +import org.apache.kafka.common.requests.FetchMetadata; +import org.apache.kafka.common.requests.FetchRequest; +import org.apache.kafka.common.requests.FetchResponse; +import org.apache.kafka.common.test.ClusterInstance; +import org.apache.kafka.common.test.api.ClusterConfigProperty; +import org.apache.kafka.common.test.api.ClusterTest; +import org.apache.kafka.common.test.api.ClusterTestDefaults; +import org.apache.kafka.common.test.api.Type; +import org.apache.kafka.common.utils.Utils; +import org.apache.kafka.common.utils.annotation.ApiKeyVersionsSource; +import org.apache.kafka.server.IntegrationTestUtils; +import org.apache.kafka.server.TestUtils; + +import org.junit.jupiter.api.Test; +import org.junit.jupiter.api.Timeout; +import org.junit.jupiter.params.ParameterizedTest; +import org.junit.jupiter.params.provider.Arguments; +import org.junit.jupiter.params.provider.MethodSource; + +import java.time.Duration; +import java.util.ArrayList; +import java.util.Collections; +import java.util.HashMap; +import java.util.HashSet; +import java.util.LinkedHashMap; +import java.util.LinkedList; +import java.util.List; +import java.util.Map; +import java.util.Optional; +import java.util.Set; +import java.util.concurrent.ExecutorService; +import java.util.concurrent.Executors; +import java.util.concurrent.Future; +import java.util.concurrent.TimeUnit; +import java.util.stream.Stream; + +import static org.apache.kafka.coordinator.group.GroupCoordinatorConfig.OFFSETS_TOPIC_REPLICATION_FACTOR_CONFIG; +import static org.apache.kafka.server.TestUtils.consumeRecords; +import static org.apache.kafka.server.config.ReplicationConfigs.REPLICA_SELECTOR_CLASS_CONFIG; +import static org.apache.kafka.server.config.ServerConfigs.BROKER_RACK_CONFIG; +import static org.apache.kafka.server.config.ServerConfigs.CONTROLLED_SHUTDOWN_ENABLE_CONFIG; +import static org.apache.kafka.server.config.ServerLogConfigs.NUM_PARTITIONS_CONFIG; +import static org.apache.kafka.test.TestUtils.waitForCondition; +import static org.junit.jupiter.api.Assertions.assertEquals; +import static org.junit.jupiter.api.Assertions.assertNotEquals; +import static org.junit.jupiter.api.Assertions.assertThrows; + +@ClusterTestDefaults( + types = {Type.KRAFT}, + brokers = FetchRequestTest.BROKER_COUNT, + serverProperties = { + @ClusterConfigProperty(id = 0, key = BROKER_RACK_CONFIG, value = "0"), + @ClusterConfigProperty(id = 1, key = BROKER_RACK_CONFIG, value = "1"), + @ClusterConfigProperty(key = REPLICA_SELECTOR_CLASS_CONFIG, + value = "org.apache.kafka.common.replica.RackAwareReplicaSelector"), + @ClusterConfigProperty(key = CONTROLLED_SHUTDOWN_ENABLE_CONFIG, value = "false"), + @ClusterConfigProperty(key = NUM_PARTITIONS_CONFIG, value = "1"), + @ClusterConfigProperty(key = OFFSETS_TOPIC_REPLICATION_FACTOR_CONFIG, value = "2"), + } +) +public class FetchRequestTest { + + public static final int BROKER_COUNT = 2; + public static final String TOPIC = "test-fetch-from-follower"; + public static final int LEADER_BROKER_ID = 0; + public static final int FOLLOWER_BROKER_ID = 1; + + private ClusterInstance cluster; + Map<String, Uuid> topicIds; + + @ClusterTest + @Timeout(15) + public void testFollowerCompleteDelayedFetchesOnReplication(ClusterInstance cluster) throws Exception { + this.cluster = cluster; + try (Admin admin = cluster.admin()) { + cluster.createTopicWithAssignment(TOPIC, Map.of(0, List.of(LEADER_BROKER_ID, FOLLOWER_BROKER_ID))); + int leaderId = TestUtils.waitUntilLeaderIsKnown(cluster.brokers().values(), new TopicPartition(TOPIC, 0)); + + var topicPartition = new TopicPartition(TOPIC, 0); + assertEquals(LEADER_BROKER_ID, leaderId); + + var fetchData = createPartitionMap(1000, List.of(topicPartition), Map.of(topicPartition, 0L)); + var fetchRequest = FetchRequest.Builder.forConsumer(ApiKeys.FETCH.latestVersion(), 20000, 1, fetchData) + .setMaxBytes(1000) + .rackId("") + .build(); + + int followerPort = cluster.brokers().get(FOLLOWER_BROKER_ID).boundPort(cluster.clientListener()); + try (var socket = IntegrationTestUtils.connect(followerPort)) { + IntegrationTestUtils.sendRequest(socket, + Utils.toArray(fetchRequest.serializeWithHeader( + IntegrationTestUtils.nextRequestHeader(ApiKeys.FETCH, ApiKeys.FETCH.latestVersion())))); + + try (Producer<byte[], byte[]> producer = cluster.producer(Map.of())) { + producer.send(new ProducerRecord<>(TOPIC, "key".getBytes(), "value".getBytes())).get(); + } + + FetchResponse response = IntegrationTestUtils.receive(socket, ApiKeys.FETCH, ApiKeys.FETCH.latestVersion()); + assertEquals(Errors.NONE, response.error()); + assertEquals(Map.of(Errors.NONE, 2), response.errorCounts()); + } + } + } + + @ClusterTest + @Timeout(15) + public void testFetchFromLeaderWhilePreferredReadReplicaIsUnavailable(ClusterInstance cluster) throws Exception { + this.cluster = cluster; + cluster.createTopicWithAssignment(TOPIC, Map.of(0, List.of(LEADER_BROKER_ID, FOLLOWER_BROKER_ID))); + TestUtils.waitUntilLeaderIsKnown(cluster.brokers().values(), new TopicPartition(TOPIC, 0)); + + produceMessages(10); + assertEquals(1, getPreferredReplica()); + + cluster.brokers().get(FOLLOWER_BROKER_ID).shutdown(); + TopicPartition topicPartition = new TopicPartition(TOPIC, 0); + + waitForCondition( + () -> { + var leaderBroker = cluster.brokers().get(LEADER_BROKER_ID); + var endpoints = leaderBroker.metadataCache() + .getPartitionReplicaEndpoints(topicPartition, cluster.clientListener()); + return !endpoints.containsKey(FOLLOWER_BROKER_ID); + }, + "follower is still reachable." + ); + + assertEquals(-1, getPreferredReplica()); + } + + @ClusterTest + @Timeout(60) + public void testFetchFromFollowerWithRoll(ClusterInstance cluster) throws Exception { + this.cluster = cluster; + cluster.createTopicWithAssignment(TOPIC, Map.of(0, List.of(LEADER_BROKER_ID, FOLLOWER_BROKER_ID))); + TestUtils.waitUntilLeaderIsKnown(cluster.brokers().values(), new TopicPartition(TOPIC, 0)); + + var consumerProps = new HashMap<String, Object>(); + consumerProps.put(ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG, cluster.bootstrapServers()); + consumerProps.put(ConsumerConfig.GROUP_ID_CONFIG, "test-group"); + consumerProps.put(ConsumerConfig.AUTO_OFFSET_RESET_CONFIG, "earliest"); + consumerProps.put(ConsumerConfig.CLIENT_RACK_CONFIG, String.valueOf(FOLLOWER_BROKER_ID)); + + try (var consumer = cluster.consumer(consumerProps)) { + consumer.subscribe(List.of(TOPIC)); + + // Wait until preferred replica is set to follower. + waitForCondition( + () -> getPreferredReplica() == FOLLOWER_BROKER_ID, + "Preferred replica is not set to follower" + ); + + // Produce and consume. + produceMessages(1); + consumeRecords(consumer, 1); + + // Shutdown follower, produce and consume should work. + cluster.shutdownBroker(FOLLOWER_BROKER_ID); + produceMessages(1); + consumeRecords(consumer, 1); + + // Start the follower and wait until preferred replica is set to follower. + cluster.startBroker(FOLLOWER_BROKER_ID); + waitForCondition( + () -> getPreferredReplica() == FOLLOWER_BROKER_ID, + "Preferred replica is not set to follower after restart" + ); + + // Produce and consume should still work. + produceMessages(1); + consumeRecords(consumer, 1); + } + } + + @ClusterTest + @Timeout(60) + public void testRackAwareRangeAssignor(ClusterInstance cluster) throws Exception { + this.cluster = cluster; + var partitionList = cluster.brokers().keySet().stream() + .sorted() + .toList(); + + String topicWithAllPartitionsOnAllRacks = "topicWithAllPartitionsOnAllRacks"; + cluster.createTopic(topicWithAllPartitionsOnAllRacks, + cluster.brokers().size(), + (short) cluster.brokers().size()); + + // Racks are in order of broker ids, assign leaders in reverse order + String topicWithSingleRackPartitions = "topicWithSingleRackPartitions"; + var replicaAssignment = new LinkedHashMap<Integer, List<Integer>>(); + for (int i : partitionList) { + replicaAssignment.put(i, List.of(cluster.brokers().size() - i - 1)); + } + cluster.createTopicWithAssignment(topicWithSingleRackPartitions, replicaAssignment); + + var consumerConfigs = cluster.brokers().keySet().stream() + .sorted() + .map(brokerId -> { + var config = new HashMap<String, Object>(); + config.put(ConsumerConfig.BOOTSTRAP_SERVERS_CONFIG, cluster.bootstrapServers()); + config.put(ConsumerConfig.GROUP_ID_CONFIG, "rack-aware-group"); + config.put(ConsumerConfig.PARTITION_ASSIGNMENT_STRATEGY_CONFIG, RangeAssignor.class.getName()); + config.put(ConsumerConfig.AUTO_OFFSET_RESET_CONFIG, "earliest"); + config.put(ConsumerConfig.CLIENT_RACK_CONFIG, String.valueOf(cluster.brokers().get(brokerId).config().rack().orElse(null))); + config.put(ConsumerConfig.GROUP_INSTANCE_ID_CONFIG, "instance-" + brokerId); + config.put(ConsumerConfig.METADATA_MAX_AGE_CONFIG, "1000"); + config.put(ConsumerConfig.ENABLE_AUTO_COMMIT_CONFIG, "false"); + return config; + }) + .toList(); + + var consumers = consumerConfigs.stream() + .map(cluster::consumer) + .toList(); + + try (Producer<byte[], byte[]> producer = cluster.producer(Map.of())) { + ExecutorService executor = Executors.newFixedThreadPool(consumers.size()); + try { + // Rack-based assignment results in partitions assigned in reverse order since partition racks are in the reverse order. + verifyRackAwareAssignments(executor, consumers, producer, partitionList, topicWithSingleRackPartitions, reverseList(partitionList)); + // Non-rack-aware assignment results in ordered partitions. + verifyRackAwareAssignments(executor, consumers, producer, partitionList, topicWithAllPartitionsOnAllRacks, partitionList); + // Rack-aware assignment with co-partitioning results in reverse assignment for both topics. + verifyRackAwareAssignments(executor, consumers, producer, partitionList, topicWithAllPartitionsOnAllRacks, reverseList(partitionList), topicWithSingleRackPartitions); + + // Perform reassignment for topicWithSingleRackPartitions to reverse the replica racks and + // verify that change in replica racks results in re-assignment based on new racks. + try (Admin admin = cluster.admin()) { + var reassignments = new HashMap<TopicPartition, Optional<NewPartitionReassignment>>(); + for (int p : partitionList) { + var newAssignment = new NewPartitionReassignment(List.of(p)); + reassignments.put(new TopicPartition(topicWithSingleRackPartitions, p), Optional.of(newAssignment)); + } + admin.alterPartitionReassignments(reassignments).all().get(30, TimeUnit.SECONDS); + } + + verifyRackAwareAssignments(executor, consumers, producer, partitionList, topicWithAllPartitionsOnAllRacks, partitionList, topicWithSingleRackPartitions); + + } finally { + executor.shutdown(); + for (Object c : consumers) { + @SuppressWarnings("unchecked") + KafkaConsumer<byte[], byte[]> consumer = (KafkaConsumer<byte[], byte[]>) c; + consumer.close(); + } + } + } + } + + private Map<String, Uuid> getTopicIds() throws Exception { + if (topicIds != null) { + return topicIds; + } + this.topicIds = new HashMap<>(); + try (Admin admin = cluster.admin()) { + var descriptions = admin.describeTopics(List.of(TOPIC)) + .allTopicNames() + .get(); + descriptions.forEach((name, desc) -> topicIds.put(name, desc.topicId())); + return topicIds; + } + } + + private LinkedHashMap<TopicPartition, FetchRequest.PartitionData> createPartitionMap( + int maxPartitionBytes, + List<TopicPartition> topicPartitions, + Map<TopicPartition, Long> offsetMap) throws Exception { + + var partitionMap = new LinkedHashMap<TopicPartition, FetchRequest.PartitionData>(); + for (TopicPartition tp : topicPartitions) { + Uuid topicId = getTopicIds().getOrDefault(tp.topic(), Uuid.ZERO_UUID); + long fetchOffset = offsetMap.getOrDefault(tp, 0L); + partitionMap.put(tp, new FetchRequest.PartitionData( + topicId, + fetchOffset, + 0L, + maxPartitionBytes, + Optional.empty() + )); + } + return partitionMap; + } + + private int getPreferredReplica() throws Exception { + var topicPartition = new TopicPartition(TOPIC, 0); + var offsetMap = Map.of(topicPartition, 0L); + var fetchData = createPartitionMap(1000, List.of(topicPartition), offsetMap); + var fetchRequest = FetchRequest.Builder.forConsumer(ApiKeys.FETCH.latestVersion(), 500, 1, fetchData) + .setMaxBytes(1000) + .rackId(String.valueOf(FOLLOWER_BROKER_ID)) + .build(); + + int leaderPort = cluster.brokers().get(LEADER_BROKER_ID).boundPort(cluster.clientListener()); + try (var socket = IntegrationTestUtils.connect(leaderPort)) { + IntegrationTestUtils.sendRequest(socket, + Utils.toArray( + fetchRequest.serializeWithHeader( + IntegrationTestUtils.nextRequestHeader(ApiKeys.FETCH, ApiKeys.FETCH.latestVersion())))); + + FetchResponse response = IntegrationTestUtils.receive(socket, ApiKeys.FETCH, ApiKeys.FETCH.latestVersion()); + assertEquals(Errors.NONE, response.error()); + assertEquals(Map.of(Errors.NONE, 2), response.errorCounts()); + assertEquals(1, response.data().responses().size()); + var topicResponse = response.data().responses().get(0); + assertEquals(1, topicResponse.partitions().size()); + return topicResponse.partitions().get(0).preferredReadReplica(); + } + } + + private void produceMessages(int numMessages) throws Exception { + try (Producer<byte[], byte[]> producer = cluster.producer(Map.of())) { + for (int i = 0; i < numMessages; i++) { + producer.send(new ProducerRecord<>(TOPIC, ("key-" + i).getBytes(), ("value-" + i).getBytes())).get(); + } + } + } + + private void verifyRackAwareAssignments( + ExecutorService executor, + List<?> consumers, + Producer<byte[], byte[]> producer, + List<Integer> partitionList, + String topic, + List<Integer> expectedPartitionOrder, + String... additionalTopics) throws Exception { + + List<String> topics = new ArrayList<>(); + topics.add(topic); + topics.addAll(List.of(additionalTopics)); + + for (Object c : consumers) { Review Comment: Done. -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
