lucasbru commented on code in PR #23670: URL: https://github.com/apache/kafka/pull/23670#discussion_r4164700774
########## core/src/test/java/kafka/server/FailingTopologyDescriptionPlugin.java: ########## @@ -0,0 +1,111 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package kafka.server; + +import org.apache.kafka.coordinator.group.api.streams.StreamsGroupTopologyDescription; +import org.apache.kafka.coordinator.group.api.streams.StreamsGroupTopologyDescriptionPlugin; +import org.apache.kafka.coordinator.group.api.streams.StreamsTopologyDescriptionPermanentFailureException; +import org.apache.kafka.coordinator.group.api.streams.StreamsTopologyDescriptionTransientFailureException; + +import java.util.Map; +import java.util.concurrent.CompletableFuture; +import java.util.concurrent.ConcurrentHashMap; +import java.util.concurrent.atomic.AtomicReference; + +/** + * Test-only {@link StreamsGroupTopologyDescriptionPlugin} whose {@code setTopology} and + * {@code deleteTopology} failures are toggled at runtime, so broker integration tests can + * exercise permanent/transient plugin-failure surfacing without a real backend. Calls that + * are not configured to fail are delegated to an in-memory store, like + * {@link InMemoryTopologyDescriptionPlugin}. + * + * <p>The broker, not the test, creates the plugin instance, so the failure toggles are + * static. Callers must {@link #reset()} before relying on this plugin, since the JVM-wide + * state otherwise leaks across test methods sharing this class. + */ +public class FailingTopologyDescriptionPlugin implements StreamsGroupTopologyDescriptionPlugin { + + public enum SetTopologyFailureMode { + NONE, PERMANENT, TRANSIENT + } + + private static final AtomicReference<SetTopologyFailureMode> SET_TOPOLOGY_FAILURE_MODE = + new AtomicReference<>(SetTopologyFailureMode.NONE); + private static final AtomicReference<RuntimeException> DELETE_TOPOLOGY_FAILURE = + new AtomicReference<>(null); + + private record Entry(int topologyEpoch, StreamsGroupTopologyDescription description) { } + + private final ConcurrentHashMap<String, Entry> store = new ConcurrentHashMap<>(); + + public static void failNextSetTopology(SetTopologyFailureMode mode) { + SET_TOPOLOGY_FAILURE_MODE.set(mode); + } + + public static void failDeleteTopologyWith(RuntimeException exception) { + DELETE_TOPOLOGY_FAILURE.set(exception); + } + + public static void reset() { + SET_TOPOLOGY_FAILURE_MODE.set(SetTopologyFailureMode.NONE); + DELETE_TOPOLOGY_FAILURE.set(null); + } + + @Override + public void configure(Map<String, ?> configs) { + // No-op. + } + + @Override + public CompletableFuture<Void> setTopology(String groupId, int topologyEpoch, StreamsGroupTopologyDescription description) { + switch (SET_TOPOLOGY_FAILURE_MODE.get()) { Review Comment: Fixed - setTopology now consumes the configured mode via getAndSet(NONE), so it fires once and reverts automatically instead of staying armed until reset(). ########## core/src/test/java/kafka/server/FailingTopologyDescriptionPlugin.java: ########## @@ -0,0 +1,111 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package kafka.server; + +import org.apache.kafka.coordinator.group.api.streams.StreamsGroupTopologyDescription; +import org.apache.kafka.coordinator.group.api.streams.StreamsGroupTopologyDescriptionPlugin; +import org.apache.kafka.coordinator.group.api.streams.StreamsTopologyDescriptionPermanentFailureException; +import org.apache.kafka.coordinator.group.api.streams.StreamsTopologyDescriptionTransientFailureException; + +import java.util.Map; +import java.util.concurrent.CompletableFuture; +import java.util.concurrent.ConcurrentHashMap; +import java.util.concurrent.atomic.AtomicReference; + +/** + * Test-only {@link StreamsGroupTopologyDescriptionPlugin} whose {@code setTopology} and + * {@code deleteTopology} failures are toggled at runtime, so broker integration tests can + * exercise permanent/transient plugin-failure surfacing without a real backend. Calls that + * are not configured to fail are delegated to an in-memory store, like + * {@link InMemoryTopologyDescriptionPlugin}. Review Comment: Fixed - fully qualified the link. ########## core/src/test/scala/unit/kafka/server/StreamsGroupTopologyDescriptionRequestTest.scala: ########## @@ -321,6 +321,258 @@ class StreamsGroupTopologyDescriptionRequestTest(cluster: ClusterInstance) exten } } + @ClusterTest(serverProperties = Array( + new ClusterConfigProperty( + key = GroupCoordinatorConfig.STREAMS_GROUP_TOPOLOGY_DESCRIPTION_PLUGIN_CLASS_CONFIG, + value = "kafka.server.FailingTopologyDescriptionPlugin") + )) + def testStreamsGroupTopologyDescriptionUpdatePermanentFailureRatchetsFailedEpoch(): Unit = { + val admin = cluster.admin() + val groupId = "test-group" + val memberId = "test-member" + val topicName = "test-topic" + + try { + FailingTopologyDescriptionPlugin.reset() + TestUtils.createOffsetsTopicWithAdmin( + admin = admin, + brokers = cluster.brokers.values().asScala.toSeq, + controllers = cluster.controllers().values().asScala.toSeq + ) + TestUtils.createTopicWithAdmin( + admin = admin, + brokers = cluster.brokers.values().asScala.toSeq, + controllers = cluster.controllers().values().asScala.toSeq, + topic = topicName, + numPartitions = 3 + ) + + // Join the group and wait until the broker solicits a topology description push. + var memberEpoch = 0 + TestUtils.waitUntilTrue(() => { + val response = streamsGroupHeartbeat( + groupId = groupId, + memberId = memberId, + rebalanceTimeoutMs = 1000, + activeTasks = List.empty, + standbyTasks = List.empty, + warmupTasks = List.empty, + topology = createMockTopology(topicName) + ) + memberEpoch = response.memberEpoch() + response.errorCode == Errors.NONE.code() && response.topologyDescriptionRequired() + }, "Broker did not solicit a topology description push within the timeout period.") + + FailingTopologyDescriptionPlugin.failNextSetTopology(FailingTopologyDescriptionPlugin.SetTopologyFailureMode.PERMANENT) + + val updateResponse = streamsGroupTopologyDescriptionUpdate( + groupId = groupId, + memberId = memberId, + topologyEpoch = topologyEpoch, + topologyDescription = createTopologyDescription(topicName) + ) + assertEquals(Errors.STREAMS_TOPOLOGY_DESCRIPTION_UPDATE_FAILED.code(), updateResponse.errorCode()) + assertEquals("topology rejected by test plugin", updateResponse.errorMessage()) + + // A permanent failure ratchets the group's failed topology epoch: the broker must not + // re-solicit another push at the same epoch. + val heartbeatAfterFailure = streamsGroupHeartbeat( + groupId = groupId, + memberId = memberId, + memberEpoch = memberEpoch, + rebalanceTimeoutMs = 1000, + activeTasks = List.empty, + standbyTasks = List.empty, + warmupTasks = List.empty + ) + assertFalse(heartbeatAfterFailure.topologyDescriptionRequired(), + "Broker must not re-solicit a topology description push once the epoch is marked permanently failed.") + + // Nothing was ever stored for this epoch. + val describedGroup = streamsGroupDescribe( + groupIds = List(groupId), + includeTopologyDescription = true + ).head + assertNotEquals(StreamsGroupDescribeResponse.TOPOLOGY_DESCRIPTION_STATUS_AVAILABLE, describedGroup.topologyDescriptionStatus()) + } finally { + FailingTopologyDescriptionPlugin.reset() + admin.close() + } + } + + @ClusterTest(serverProperties = Array( + new ClusterConfigProperty( + key = GroupCoordinatorConfig.STREAMS_GROUP_TOPOLOGY_DESCRIPTION_PLUGIN_CLASS_CONFIG, + value = "kafka.server.FailingTopologyDescriptionPlugin") + )) + def testStreamsGroupTopologyDescriptionUpdateTransientFailureArmsBackOff(): Unit = { + val admin = cluster.admin() + val groupId = "test-group" + val memberId = "test-member" + val topicName = "test-topic" + + try { + FailingTopologyDescriptionPlugin.reset() + TestUtils.createOffsetsTopicWithAdmin( + admin = admin, + brokers = cluster.brokers.values().asScala.toSeq, + controllers = cluster.controllers().values().asScala.toSeq + ) + TestUtils.createTopicWithAdmin( + admin = admin, + brokers = cluster.brokers.values().asScala.toSeq, + controllers = cluster.controllers().values().asScala.toSeq, + topic = topicName, + numPartitions = 3 + ) + + // Join the group and wait until the broker solicits a topology description push. + var memberEpoch = 0 + TestUtils.waitUntilTrue(() => { + val response = streamsGroupHeartbeat( + groupId = groupId, + memberId = memberId, + rebalanceTimeoutMs = 1000, + activeTasks = List.empty, + standbyTasks = List.empty, + warmupTasks = List.empty, + topology = createMockTopology(topicName) + ) + memberEpoch = response.memberEpoch() + response.errorCode == Errors.NONE.code() && response.topologyDescriptionRequired() + }, "Broker did not solicit a topology description push within the timeout period.") + + FailingTopologyDescriptionPlugin.failNextSetTopology(FailingTopologyDescriptionPlugin.SetTopologyFailureMode.TRANSIENT) + + val updateResponse = streamsGroupTopologyDescriptionUpdate( + groupId = groupId, + memberId = memberId, + topologyEpoch = topologyEpoch, + topologyDescription = createTopologyDescription(topicName) + ) + assertEquals(Errors.STREAMS_TOPOLOGY_DESCRIPTION_UPDATE_FAILED.code(), updateResponse.errorCode()) + assertEquals("backend offline", updateResponse.errorMessage()) + + // A transient failure arms the per-group back-off: a heartbeat immediately afterwards + // must not re-solicit another push. + val heartbeatAfterFailure = streamsGroupHeartbeat( + groupId = groupId, + memberId = memberId, + memberEpoch = memberEpoch, + rebalanceTimeoutMs = 1000, + activeTasks = List.empty, + standbyTasks = List.empty, + warmupTasks = List.empty + ) + assertFalse(heartbeatAfterFailure.topologyDescriptionRequired(), + "Broker must not re-solicit a topology description push immediately after a transient failure; back-off must be armed.") Review Comment: Fair - the assertion was circular, the window armed by the original solicitation alone explains the suppression. Rewrote the test to sleep out that window (worst case 36s) before triggering the failure, so the suppression after can only come from the failure path, then waits for the back-off to lapse and the broker to re-solicit at the same epoch. ########## core/src/test/scala/unit/kafka/server/StreamsGroupTopologyDescriptionRequestTest.scala: ########## @@ -321,6 +321,258 @@ class StreamsGroupTopologyDescriptionRequestTest(cluster: ClusterInstance) exten } } + @ClusterTest(serverProperties = Array( + new ClusterConfigProperty( + key = GroupCoordinatorConfig.STREAMS_GROUP_TOPOLOGY_DESCRIPTION_PLUGIN_CLASS_CONFIG, + value = "kafka.server.FailingTopologyDescriptionPlugin") + )) + def testStreamsGroupTopologyDescriptionUpdatePermanentFailureRatchetsFailedEpoch(): Unit = { + val admin = cluster.admin() + val groupId = "test-group" + val memberId = "test-member" + val topicName = "test-topic" + + try { + FailingTopologyDescriptionPlugin.reset() + TestUtils.createOffsetsTopicWithAdmin( + admin = admin, + brokers = cluster.brokers.values().asScala.toSeq, + controllers = cluster.controllers().values().asScala.toSeq + ) + TestUtils.createTopicWithAdmin( + admin = admin, + brokers = cluster.brokers.values().asScala.toSeq, + controllers = cluster.controllers().values().asScala.toSeq, + topic = topicName, + numPartitions = 3 + ) + + // Join the group and wait until the broker solicits a topology description push. + var memberEpoch = 0 + TestUtils.waitUntilTrue(() => { + val response = streamsGroupHeartbeat( + groupId = groupId, + memberId = memberId, + rebalanceTimeoutMs = 1000, + activeTasks = List.empty, + standbyTasks = List.empty, + warmupTasks = List.empty, + topology = createMockTopology(topicName) + ) + memberEpoch = response.memberEpoch() + response.errorCode == Errors.NONE.code() && response.topologyDescriptionRequired() + }, "Broker did not solicit a topology description push within the timeout period.") + + FailingTopologyDescriptionPlugin.failNextSetTopology(FailingTopologyDescriptionPlugin.SetTopologyFailureMode.PERMANENT) + + val updateResponse = streamsGroupTopologyDescriptionUpdate( + groupId = groupId, + memberId = memberId, + topologyEpoch = topologyEpoch, + topologyDescription = createTopologyDescription(topicName) + ) + assertEquals(Errors.STREAMS_TOPOLOGY_DESCRIPTION_UPDATE_FAILED.code(), updateResponse.errorCode()) + assertEquals("topology rejected by test plugin", updateResponse.errorMessage()) + + // A permanent failure ratchets the group's failed topology epoch: the broker must not + // re-solicit another push at the same epoch. + val heartbeatAfterFailure = streamsGroupHeartbeat( + groupId = groupId, + memberId = memberId, + memberEpoch = memberEpoch, + rebalanceTimeoutMs = 1000, + activeTasks = List.empty, + standbyTasks = List.empty, + warmupTasks = List.empty + ) + assertFalse(heartbeatAfterFailure.topologyDescriptionRequired(), + "Broker must not re-solicit a topology description push once the epoch is marked permanently failed.") + + // Nothing was ever stored for this epoch. + val describedGroup = streamsGroupDescribe( + groupIds = List(groupId), + includeTopologyDescription = true + ).head + assertNotEquals(StreamsGroupDescribeResponse.TOPOLOGY_DESCRIPTION_STATUS_AVAILABLE, describedGroup.topologyDescriptionStatus()) + } finally { + FailingTopologyDescriptionPlugin.reset() + admin.close() + } + } + + @ClusterTest(serverProperties = Array( + new ClusterConfigProperty( + key = GroupCoordinatorConfig.STREAMS_GROUP_TOPOLOGY_DESCRIPTION_PLUGIN_CLASS_CONFIG, + value = "kafka.server.FailingTopologyDescriptionPlugin") + )) + def testStreamsGroupTopologyDescriptionUpdateTransientFailureArmsBackOff(): Unit = { + val admin = cluster.admin() + val groupId = "test-group" + val memberId = "test-member" + val topicName = "test-topic" + + try { + FailingTopologyDescriptionPlugin.reset() + TestUtils.createOffsetsTopicWithAdmin( + admin = admin, + brokers = cluster.brokers.values().asScala.toSeq, + controllers = cluster.controllers().values().asScala.toSeq + ) + TestUtils.createTopicWithAdmin( + admin = admin, + brokers = cluster.brokers.values().asScala.toSeq, + controllers = cluster.controllers().values().asScala.toSeq, + topic = topicName, + numPartitions = 3 + ) + + // Join the group and wait until the broker solicits a topology description push. + var memberEpoch = 0 + TestUtils.waitUntilTrue(() => { + val response = streamsGroupHeartbeat( + groupId = groupId, + memberId = memberId, + rebalanceTimeoutMs = 1000, + activeTasks = List.empty, + standbyTasks = List.empty, + warmupTasks = List.empty, + topology = createMockTopology(topicName) + ) + memberEpoch = response.memberEpoch() + response.errorCode == Errors.NONE.code() && response.topologyDescriptionRequired() + }, "Broker did not solicit a topology description push within the timeout period.") + + FailingTopologyDescriptionPlugin.failNextSetTopology(FailingTopologyDescriptionPlugin.SetTopologyFailureMode.TRANSIENT) + + val updateResponse = streamsGroupTopologyDescriptionUpdate( + groupId = groupId, + memberId = memberId, + topologyEpoch = topologyEpoch, + topologyDescription = createTopologyDescription(topicName) + ) + assertEquals(Errors.STREAMS_TOPOLOGY_DESCRIPTION_UPDATE_FAILED.code(), updateResponse.errorCode()) + assertEquals("backend offline", updateResponse.errorMessage()) + + // A transient failure arms the per-group back-off: a heartbeat immediately afterwards + // must not re-solicit another push. + val heartbeatAfterFailure = streamsGroupHeartbeat( + groupId = groupId, + memberId = memberId, + memberEpoch = memberEpoch, + rebalanceTimeoutMs = 1000, + activeTasks = List.empty, + standbyTasks = List.empty, + warmupTasks = List.empty + ) + assertFalse(heartbeatAfterFailure.topologyDescriptionRequired(), + "Broker must not re-solicit a topology description push immediately after a transient failure; back-off must be armed.") Review Comment: Done, same fix as the Copilot thread above - test now waits out the initial window with real time, then waits for the back-off to lapse and verifies re-solicitation at the same epoch. -- 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]
