mimaison commented on code in PR #21689: URL: https://github.com/apache/kafka/pull/21689#discussion_r2917209597
########## server/src/test/java/org/apache/kafka/server/UncleanLeaderElectionTest.java: ########## @@ -0,0 +1,466 @@ +/* + * 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; + +import kafka.server.KafkaBroker; + +import org.apache.kafka.clients.admin.Admin; +import org.apache.kafka.clients.admin.AlterConfigOp; +import org.apache.kafka.clients.admin.AlterConfigsResult; +import org.apache.kafka.clients.admin.ConfigEntry; +import org.apache.kafka.clients.admin.FeatureUpdate; +import org.apache.kafka.clients.admin.NewTopic; +import org.apache.kafka.clients.consumer.Consumer; +import org.apache.kafka.clients.consumer.ConsumerConfig; +import org.apache.kafka.clients.consumer.ConsumerRecord; +import org.apache.kafka.clients.consumer.GroupProtocol; +import org.apache.kafka.common.TopicPartition; +import org.apache.kafka.common.config.ConfigResource; +import org.apache.kafka.common.config.TopicConfig; +import org.apache.kafka.common.errors.InvalidConfigurationException; +import org.apache.kafka.common.errors.TimeoutException; +import org.apache.kafka.common.serialization.StringDeserializer; +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.utils.Utils; +import org.apache.kafka.controller.ReplicationControlManager; +import org.apache.kafka.coordinator.group.GroupCoordinatorConfig; +import org.apache.kafka.metadata.LeaderAndIsr; +import org.apache.kafka.metadata.LeaderConstants; +import org.apache.kafka.metadata.MetadataCache; +import org.apache.kafka.server.common.EligibleLeaderReplicasVersion; +import org.apache.kafka.server.config.ReplicationConfigs; +import org.apache.kafka.server.metrics.KafkaYammerMetrics; + +import com.yammer.metrics.core.Meter; + +import org.apache.logging.log4j.Level; +import org.apache.logging.log4j.core.config.Configurator; +import org.junit.jupiter.api.AfterEach; +import org.junit.jupiter.api.BeforeEach; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import java.util.Collection; +import java.util.List; +import java.util.Map; +import java.util.Optional; +import java.util.Random; +import java.util.Set; +import java.util.concurrent.ExecutionException; + +import static org.apache.kafka.server.TestUtils.awaitLeaderChange; +import static org.apache.kafka.server.TestUtils.produceMessage; +import static org.apache.kafka.server.TestUtils.waitForPartitionMetadata; +import static org.apache.kafka.test.TestUtils.DEFAULT_MAX_WAIT_MS; +import static org.apache.kafka.test.TestUtils.waitForCondition; +import static org.junit.jupiter.api.Assertions.assertEquals; +import static org.junit.jupiter.api.Assertions.assertInstanceOf; +import static org.junit.jupiter.api.Assertions.assertThrows; +import static org.junit.jupiter.api.Assertions.assertTrue; + +@ClusterTestDefaults( + brokers = 2, + serverProperties = { + // trim the retry count and backoff interval to reduce test execution time + @ClusterConfigProperty(key = ReplicationConfigs.UNCLEAN_LEADER_ELECTION_INTERVAL_MS_CONFIG, value = "10"), + @ClusterConfigProperty(key = GroupCoordinatorConfig.OFFSETS_TOPIC_REPLICATION_FACTOR_CONFIG, value = "1") + } +) +public class UncleanLeaderElectionTest { + + private static final Logger LOG = LoggerFactory.getLogger(UncleanLeaderElectionTest.class); + + private static final int BROKER_ID_0 = 0; + private static final int BROKER_ID_1 = 1; + private static final Random RANDOM = new Random(); + private static final String TOPIC = "topic" + RANDOM.nextLong(); + private static final int PARTITION_ID = 0; + private static final TopicPartition TOPIC_PARTITION = new TopicPartition(TOPIC, PARTITION_ID); + + private final ClusterInstance cluster; + private Admin admin; + + public UncleanLeaderElectionTest(ClusterInstance cluster) { + this.cluster = cluster; + } + + @BeforeEach + public void setup() { + admin = cluster.admin(); + // temporarily set ReplicationControlManager logger to a higher level so that tests run quietly + Configurator.setLevel(ReplicationControlManager.class.getName(), Level.FATAL); + } + + @AfterEach + public void teardown() { + Utils.closeQuietly(admin, "admin client"); + // restore log level + Configurator.setLevel(ReplicationControlManager.class.getName(), Level.ERROR); + } + + private void disableEligibleLeaderReplicas() throws Exception { + admin.updateFeatures( + Map.of(EligibleLeaderReplicasVersion.FEATURE_NAME, new FeatureUpdate((short) 0, FeatureUpdate.UpgradeType.SAFE_DOWNGRADE)) + ).all().get(); + } + + @ClusterTest( + serverProperties = { + @ClusterConfigProperty(key = TopicConfig.UNCLEAN_LEADER_ELECTION_ENABLE_CONFIG, value = "true") + } + ) + public void testUncleanLeaderElectionEnabledClassic() throws Exception { + testUncleanLeaderElectionEnabled(GroupProtocol.CLASSIC); + } + + @ClusterTest( + serverProperties = { + @ClusterConfigProperty(key = TopicConfig.UNCLEAN_LEADER_ELECTION_ENABLE_CONFIG, value = "true") + } + ) + public void testUncleanLeaderElectionEnabledConsumer() throws Exception { + testUncleanLeaderElectionEnabled(GroupProtocol.CONSUMER); + } + + private void testUncleanLeaderElectionEnabled(GroupProtocol groupProtocol) throws Exception { + disableEligibleLeaderReplicas(); + + // create topic with 1 partition, 2 replicas, one on each broker + NewTopic newTopic = new NewTopic(TOPIC, Map.of(PARTITION_ID, List.of(BROKER_ID_0, BROKER_ID_1))); + admin.createTopics(List.of(newTopic)).all().get(); + + verifyUncleanLeaderElectionEnabled(groupProtocol); + } + + @ClusterTest + public void testUncleanLeaderElectionDisabledClassic() throws Exception { + testUncleanLeaderElectionDisabled(GroupProtocol.CLASSIC); + } + + @ClusterTest + public void testUncleanLeaderElectionDisabledConsumer() throws Exception { + testUncleanLeaderElectionDisabled(GroupProtocol.CONSUMER); + } + + private void testUncleanLeaderElectionDisabled(GroupProtocol groupProtocol) throws Exception { + // unclean leader election is disabled by default + disableEligibleLeaderReplicas(); + + // create topic with 1 partition, 2 replicas, one on each broker + NewTopic newTopic = new NewTopic(TOPIC, Map.of(PARTITION_ID, List.of(BROKER_ID_0, BROKER_ID_1))); + admin.createTopics(List.of(newTopic)).all().get(); + + verifyUncleanLeaderElectionDisabled(groupProtocol); + } + + @ClusterTest + public void testUncleanLeaderElectionEnabledByTopicOverrideClassic() throws Exception { + testUncleanLeaderElectionEnabledByTopicOverride(GroupProtocol.CLASSIC); + } + + @ClusterTest + public void testUncleanLeaderElectionEnabledByTopicOverrideConsumer() throws Exception { + testUncleanLeaderElectionEnabledByTopicOverride(GroupProtocol.CONSUMER); + } + + private void testUncleanLeaderElectionEnabledByTopicOverride(GroupProtocol groupProtocol) throws Exception { + disableEligibleLeaderReplicas(); + + // create topic with 1 partition, 2 replicas, one on each broker, and unclean leader election enabled + NewTopic newTopic = new NewTopic(TOPIC, Map.of(PARTITION_ID, List.of(BROKER_ID_0, BROKER_ID_1))) + .configs(Map.of(TopicConfig.UNCLEAN_LEADER_ELECTION_ENABLE_CONFIG, "true")); + admin.createTopics(List.of(newTopic)).all().get(); + + verifyUncleanLeaderElectionEnabled(groupProtocol); + } + + @ClusterTest( + serverProperties = { + @ClusterConfigProperty(key = TopicConfig.UNCLEAN_LEADER_ELECTION_ENABLE_CONFIG, value = "true") + } + ) + public void testUncleanLeaderElectionDisabledByTopicOverrideClassic() throws Exception { + testUncleanLeaderElectionDisabledByTopicOverride(GroupProtocol.CLASSIC); + } + + @ClusterTest( + serverProperties = { + @ClusterConfigProperty(key = TopicConfig.UNCLEAN_LEADER_ELECTION_ENABLE_CONFIG, value = "true") + } + ) + public void testUncleanLeaderElectionDisabledByTopicOverrideConsumer() throws Exception { + testUncleanLeaderElectionDisabledByTopicOverride(GroupProtocol.CONSUMER); + } + + private void testUncleanLeaderElectionDisabledByTopicOverride(GroupProtocol groupProtocol) throws Exception { + disableEligibleLeaderReplicas(); + + // create topic with 1 partition, 2 replicas, one on each broker, and unclean leader election disabled + NewTopic newTopic = new NewTopic(TOPIC, Map.of(PARTITION_ID, List.of(BROKER_ID_0, BROKER_ID_1))) + .configs(Map.of(TopicConfig.UNCLEAN_LEADER_ELECTION_ENABLE_CONFIG, "false")); + admin.createTopics(List.of(newTopic)).all().get(); + + verifyUncleanLeaderElectionDisabled(groupProtocol); + } + + @ClusterTest + public void testUncleanLeaderElectionInvalidTopicOverride() throws Exception { + disableEligibleLeaderReplicas(); + + // create topic with an invalid value for unclean leader election + NewTopic newTopic = new NewTopic(TOPIC, Map.of(PARTITION_ID, List.of(BROKER_ID_0, BROKER_ID_1))) + .configs(Map.of(TopicConfig.UNCLEAN_LEADER_ELECTION_ENABLE_CONFIG, "invalid")); + Throwable e = assertThrows(ExecutionException.class, () -> admin.createTopics(List.of(newTopic)).all().get()); + assertInstanceOf(InvalidConfigurationException.class, e.getCause()); + } + + private void verifyUncleanLeaderElectionEnabled(GroupProtocol groupProtocol) throws Exception { + // wait until leader is elected + int leaderId = awaitLeaderChange(cluster, TOPIC_PARTITION, Optional.empty()); + LOG.debug("Leader for {} is elected to be: {}", TOPIC, leaderId); + assertTrue(leaderId == BROKER_ID_0 || leaderId == BROKER_ID_1, + "Leader id is set to expected value for topic: " + TOPIC); + + // the non-leader broker is the follower + int followerId = leaderId == BROKER_ID_0 ? BROKER_ID_1 : BROKER_ID_0; + LOG.debug("Follower for {} is: {}", TOPIC, followerId); + + produceMessage(cluster, TOPIC, "first"); + waitForPartitionMetadata(cluster.brokers().values(), TOPIC, PARTITION_ID); + assertEquals(List.of("first"), consumeAllMessages(1, groupProtocol)); + + // shutdown follower broker + KafkaBroker follower = cluster.brokers().get(followerId); + follower.shutdown(); + follower.awaitShutdown(); + + produceMessage(cluster, TOPIC, "second"); + assertEquals(List.of("first", "second"), consumeAllMessages(2, groupProtocol)); + + //verify that unclean election metric count is 0 + assertEquals(0L, getLeaderElectionCount()); + + // shutdown leader and then restart follower + KafkaBroker leader = cluster.brokers().get(leaderId); + leader.shutdown(); + leader.awaitShutdown(); + follower.startup(); + + // wait until new leader is (uncleanly) elected + awaitLeaderChange(cluster, TOPIC_PARTITION, Optional.of(followerId), 30000); + assertEquals(1L, getLeaderElectionCount()); + + produceMessage(cluster, TOPIC, "third"); + + // second message was lost due to unclean election + assertEquals(List.of("first", "third"), consumeAllMessages(2, groupProtocol)); + } + + private void verifyUncleanLeaderElectionDisabled(GroupProtocol groupProtocol) throws Exception { + // wait until leader is elected + int leaderId = awaitLeaderChange(cluster, TOPIC_PARTITION, Optional.empty()); + LOG.debug("Leader for {} is elected to be: {}", TOPIC, leaderId); + assertTrue(leaderId == BROKER_ID_0 || leaderId == BROKER_ID_1, + "Leader id is set to expected value for topic: " + TOPIC); + + // the non-leader broker is the follower + int followerId = leaderId == BROKER_ID_0 ? BROKER_ID_1 : BROKER_ID_0; + LOG.debug("Follower for {} is: {}", TOPIC, followerId); + + produceMessage(cluster, TOPIC, "first"); + waitForPartitionMetadata(cluster.brokers().values(), TOPIC, PARTITION_ID); + assertEquals(List.of("first"), consumeAllMessages(1, groupProtocol)); + + // shutdown follower broker + KafkaBroker follower = cluster.brokers().get(followerId); + follower.shutdown(); + follower.awaitShutdown(); + + produceMessage(cluster, TOPIC, "second"); + assertEquals(List.of("first", "second"), consumeAllMessages(2, groupProtocol)); + + //remove any previous unclean election metric + assertEquals(0L, getLeaderElectionCount()); + + // shutdown leader and then restart follower + KafkaBroker leader = cluster.brokers().get(leaderId); + leader.shutdown(); + leader.awaitShutdown(); + follower.startup(); + + // verify that unclean election to non-ISR follower does not occur. + // That is, leader should be NO_LEADER(-1) and the ISR should have only old leaderId. + waitForNoLeaderAndIsrHasOldLeaderId(follower.replicaManager().metadataCache(), leaderId); + assertEquals(0L, getLeaderElectionCount()); + + // message production and consumption should both fail while leader is down + Throwable e = assertThrows(ExecutionException.class, () -> produceMessage(cluster, TOPIC, "third", 1000, 1000)); + assertInstanceOf(TimeoutException.class, e.getCause()); + + assertEquals(List.of(), consumeAllMessages(0, groupProtocol)); + + // restart leader temporarily to send a successfully replicated message + leader.startup(); + awaitLeaderChange(cluster, TOPIC_PARTITION, Optional.of(leaderId)); + + produceMessage(cluster, TOPIC, "third"); + //make sure follower server joins the ISR + waitForCondition( + () -> { + Optional<LeaderAndIsr> partitionInfoOpt = follower.metadataCache().getLeaderAndIsr(TOPIC, PARTITION_ID); + return partitionInfoOpt.isPresent() && partitionInfoOpt.get().isr().contains(followerId); + }, DEFAULT_MAX_WAIT_MS, "Inconsistent metadata after first server startup"); + + leader.shutdown(); + leader.awaitShutdown(); + + // verify clean leader transition to ISR follower + awaitLeaderChange(cluster, TOPIC_PARTITION, Optional.of(followerId)); + // verify messages can be consumed from ISR follower that was just promoted to leader + assertEquals(List.of("first", "second", "third"), consumeAllMessages(3, groupProtocol)); + } + + private long getLeaderElectionCount() { + Meter meter = (Meter) KafkaYammerMetrics.defaultRegistry().allMetrics().entrySet().stream() + .filter(entry -> entry.getKey().getName().endsWith("UncleanLeaderElectionsPerSec")) + .map(Map.Entry::getValue) + .findFirst() + .orElseThrow(() -> new AssertionError("Unable to find metric UncleanLeaderElectionsPerSec")); + return meter.count(); + } + + private List<String> consumeAllMessages(int numMessages, GroupProtocol groupProtocol) throws Exception { + Map<String, Object> consumerConfig = Map.of( + ConsumerConfig.KEY_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class.getName(), + ConsumerConfig.VALUE_DESERIALIZER_CLASS_CONFIG, StringDeserializer.class.getName(), + ConsumerConfig.GROUP_ID_CONFIG, "group" + RANDOM.nextLong(), + ConsumerConfig.ENABLE_AUTO_COMMIT_CONFIG, false, + ConsumerConfig.GROUP_PROTOCOL_CONFIG, groupProtocol.name() + ); + + try (Consumer<String, String> consumer = cluster.consumer(consumerConfig)) { + consumer.assign(List.of(TOPIC_PARTITION)); + consumer.seekToBeginning(List.of(TOPIC_PARTITION)); + return TestUtils.consumeRecords(consumer, numMessages).stream().map(ConsumerRecord::value).toList(); + } + } + + @ClusterTest + public void testTopicUncleanLeaderElectionEnableWithAlterTopicConfigsClassic() throws Exception { + testTopicUncleanLeaderElectionEnableWithAlterTopicConfigs(GroupProtocol.CLASSIC); + } + + @ClusterTest + public void testTopicUncleanLeaderElectionEnableWithAlterTopicConfigsConsumer() throws Exception { + testTopicUncleanLeaderElectionEnableWithAlterTopicConfigs(GroupProtocol.CONSUMER); + } + + private void testTopicUncleanLeaderElectionEnableWithAlterTopicConfigs(GroupProtocol groupProtocol) throws Exception { + disableEligibleLeaderReplicas(); + + // create topic with 1 partition, 2 replicas, one on each broker + NewTopic newTopic = new NewTopic(TOPIC, Map.of(PARTITION_ID, List.of(BROKER_ID_0, BROKER_ID_1))); + admin.createTopics(List.of(newTopic)).all().get(); + + // wait until leader is elected + int leaderId = awaitLeaderChange(cluster, TOPIC_PARTITION, Optional.empty()); + + // the non-leader broker is the follower + int followerId = leaderId == BROKER_ID_0 ? BROKER_ID_1 : BROKER_ID_0; + + produceMessage(cluster, TOPIC, "first"); + waitForPartitionMetadata(cluster.brokers().values(), TOPIC, PARTITION_ID); + assertEquals(List.of("first"), consumeAllMessages(1, groupProtocol)); + + // Verify the "unclean.leader.election.enable" won't be triggered even if it is enabled/disabled dynamically, + // because the leader is still alive + Map<String, String> newProps = Map.of(ReplicationConfigs.UNCLEAN_LEADER_ELECTION_ENABLE_CONFIG, "true"); + alterTopicConfigs(newProps).all().get(); + // leader should not change to followerId + awaitLeaderChange(cluster, TOPIC_PARTITION, Optional.of(leaderId), 10000); + + newProps = Map.of(ReplicationConfigs.UNCLEAN_LEADER_ELECTION_ENABLE_CONFIG, "false"); + alterTopicConfigs(newProps).all().get(); + // leader should not change to followerId + awaitLeaderChange(cluster, TOPIC_PARTITION, Optional.of(leaderId), 10000); + + // shutdown follower server + KafkaBroker follower = cluster.brokers().get(followerId); + follower.shutdown(); + follower.awaitShutdown(); + + produceMessage(cluster, TOPIC, "second"); + assertEquals(List.of("first", "second"), consumeAllMessages(2, groupProtocol)); + + // verify that unclean election metric count is 0 + assertEquals(0L, getLeaderElectionCount()); + + // shutdown leader and then restart follower + KafkaBroker leader = cluster.brokers().get(leaderId); + leader.shutdown(); + leader.awaitShutdown(); + follower.startup(); + + // verify that unclean election to non-ISR follower does not occur. + // That is, leader should be NO_LEADER(-1) and the ISR should have only old leaderId. + waitForNoLeaderAndIsrHasOldLeaderId(follower.replicaManager().metadataCache(), leaderId); + assertEquals(0L, getLeaderElectionCount()); + + // message production and consumption should both fail while leader is down + Throwable e = assertThrows(ExecutionException.class, () -> produceMessage(cluster, TOPIC, "third", 1000, 1000)); + assertEquals(TimeoutException.class, e.getCause().getClass()); + + assertEquals(List.of(), consumeAllMessages(0, groupProtocol)); + + // Enable unclean leader election for topic + newProps = Map.of(ReplicationConfigs.UNCLEAN_LEADER_ELECTION_ENABLE_CONFIG, "true"); + alterTopicConfigs(newProps).all().get(); + + // wait until new leader is (uncleanly) elected + awaitLeaderChange(cluster, TOPIC_PARTITION, Optional.of(followerId), 30000); + assertEquals(1L, getLeaderElectionCount()); + + produceMessage(cluster, TOPIC, "third"); + + // second message was lost due to unclean election + assertEquals(List.of("first", "third"), consumeAllMessages(2, groupProtocol)); + } + + private AlterConfigsResult alterTopicConfigs(Map<String, String> configs) { + ConfigResource configResource = new ConfigResource(ConfigResource.Type.TOPIC, TOPIC); + + Collection<AlterConfigOp> configEntries = configs.entrySet().stream() Review Comment: Good idea, I pushed an update. Thanks -- 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]
