You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by "lucasbru (via GitHub)" <gi...@apache.org> on 2026/10/01 18:41:35 UTC
[PR] MINOR: Add broker integration tests for topology-description plugin failure surfacing [kafka]
lucasbru opened a new pull request, #23670:
URL: https://github.com/apache/kafka/pull/23670
StreamsGroupTopologyDescriptionRequestTest only ever exercises InMemoryTopologyDescriptionPlugin, which never fails, so the permanent and transient failure classification for the StreamsGroupTopologyDescriptionUpdate RPC, and the DeleteGroups plugin-failure path, were proven only against mocks in GroupCoordinatorServiceTopologyDescriptionTest, never against a real broker/RPC round trip.
This adds FailingTopologyDescriptionPlugin, a configurable test double whose setTopology and deleteTopology calls can be told to fail on demand, and three new cluster tests built on it. A permanent setTopology failure surfaces STREAMS_TOPOLOGY_DESCRIPTION_UPDATE_FAILED with the plugin's message and stops the broker from re-soliciting a push at that epoch. A transient failure returns the same error code but arms the heartbeat back-off instead of ratcheting the epoch. A DeleteGroups plugin failure returns GROUP_DELETION_FAILED without tombstoning the group, and a retried delete succeeds once the plugin recovers.
KSTREAMS-9351
--
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-unsubscribe@kafka.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] MINOR: Add broker integration tests for topology-description plugin failure surfacing [kafka]
Posted by "lucasbru (via GitHub)" <gi...@apache.org>.
lucasbru commented on code in PR #23670:
URL: https://github.com/apache/kafka/pull/23670#discussion_r4165529190
##########
core/src/test/scala/unit/kafka/server/StreamsGroupTopologyDescriptionRequestTest.scala:
##########
@@ -307,7 +308,7 @@ class StreamsGroupTopologyDescriptionRequestTest(cluster: ClusterInstance) exten
deleteGroups(
groupIds = List(groupId),
expectedErrors = List(Errors.NONE),
- version = ApiKeys.DELETE_GROUPS.latestVersion(isUnstableApiEnabled)
Review Comment:
not required anymore. The new version of DELETE_GROUPS is marked stable.
--
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-unsubscribe@kafka.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] MINOR: Add broker integration tests for topology-description plugin failure surfacing [kafka]
Posted by "lucasbru (via GitHub)" <gi...@apache.org>.
lucasbru commented on code in PR #23670:
URL: https://github.com/apache/kafka/pull/23670#discussion_r4209763835
##########
core/src/test/scala/unit/kafka/server/StreamsGroupTopologyDescriptionRequestTest.scala:
##########
@@ -321,6 +322,280 @@ 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
+ )
+
+ val memberEpoch = joinAndAwaitTopologyDescriptionSolicited(groupId, memberId, topicName)
+
+ 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: the permanent failure leaves the group's
+ // stored-topology-epoch at its UNCERTAIN sentinel, which Describe reports as NOT_STORED.
+ val describedGroup = streamsGroupDescribe(
+ groupIds = List(groupId),
+ includeTopologyDescription = true
+ ).head
+ assertEquals(StreamsGroupDescribeResponse.TOPOLOGY_DESCRIPTION_STATUS_NOT_STORED, describedGroup.topologyDescriptionStatus())
+ } finally {
+ FailingTopologyDescriptionPlugin.reset()
+ admin.close()
+ }
+ }
+
+ @Timeout(180)
+ @ClusterTest(serverProperties = Array(
+ new ClusterConfigProperty(
+ key = GroupCoordinatorConfig.STREAMS_GROUP_TOPOLOGY_DESCRIPTION_PLUGIN_CLASS_CONFIG,
+ value = "kafka.server.FailingTopologyDescriptionPlugin"),
+ // The test below goes 40s without heartbeating; the default 45s session timeout would
+ // leave only ~5s of margin before the member is reaped for inactivity.
+ new ClusterConfigProperty(key = GroupCoordinatorConfig.STREAMS_GROUP_SESSION_TIMEOUT_MS_CONFIG, value = "120000"),
+ new ClusterConfigProperty(key = GroupCoordinatorConfig.STREAMS_GROUP_MAX_SESSION_TIMEOUT_MS_CONFIG, value = "120000")
+ ))
+ 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. This
+ // arms the first back-off window: 30s * 2^0, +/-20% jitter (StreamsGroupTopologyDescriptionBackoff),
+ // i.e. 24-36s.
+ var memberEpoch = joinAndAwaitTopologyDescriptionSolicited(groupId, memberId, topicName)
+
+ // Let that window fully expire without heartbeating, so nothing but the upcoming
+ // transient failure can be responsible for the suppression asserted below. The back-off
+ // uses the broker's real clock, with no test seam to inject a mock one here, so this
+ // waits out real time rather than simulating it.
+ Thread.sleep(40000)
Review Comment:
Couldn't find a good way to avoid the sleep here.
--
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-unsubscribe@kafka.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] MINOR: Add broker integration tests for topology-description plugin failure surfacing [kafka]
Posted by "Copilot (via GitHub)" <gi...@apache.org>.
Copilot commented on code in PR #23670:
URL: https://github.com/apache/kafka/pull/23670#discussion_r4164795874
##########
core/src/test/scala/unit/kafka/server/StreamsGroupTopologyDescriptionRequestTest.scala:
##########
@@ -321,6 +323,282 @@ 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()
+ }
+ }
+
+ @Timeout(180)
+ @ClusterTest(serverProperties = Array(
+ new ClusterConfigProperty(
+ key = GroupCoordinatorConfig.STREAMS_GROUP_TOPOLOGY_DESCRIPTION_PLUGIN_CLASS_CONFIG,
+ value = "kafka.server.FailingTopologyDescriptionPlugin")
+ ))
Review Comment:
This test later stops heartbeating for 40 seconds, while the default Streams session timeout is only 45 seconds (`GroupCoordinatorConfig.java:343-344`). That leaves just five seconds for CI or scheduler delays before the member expires, making the update/heartbeat fail for membership reasons instead of testing transient back-off. Configure a session timeout comfortably above the sleep (and raise its maximum accordingly) for this test.
--
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-unsubscribe@kafka.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] MINOR: Add broker integration tests for topology-description plugin failure surfacing [kafka]
Posted by "andyhuangdev (via GitHub)" <gi...@apache.org>.
andyhuangdev commented on code in PR #23670:
URL: https://github.com/apache/kafka/pull/23670#discussion_r4162070343
##########
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:
Let the initial solicitation window expire before triggering the transient failure (or control the broker clock/back-off), then verify suppression after the failure and eventual re-solicitation at the same topology 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: jira-unsubscribe@kafka.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] MINOR: Add broker integration tests for topology-description plugin failure surfacing [kafka]
Posted by "Copilot (via GitHub)" <gi...@apache.org>.
Copilot commented on code in PR #23670:
URL: https://github.com/apache/kafka/pull/23670#discussion_r4159262277
##########
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:
`failNextSetTopology` does not currently fail only the next call: reading the mode with `get()` leaves it active, so every later `setTopology` call fails until an explicit `reset()`. Consume the configured failure atomically so this test double matches its API and can model a transient failure followed by recovery.
##########
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:
This Javadoc link uses an unimported simple name, but `InMemoryTopologyDescriptionPlugin` is in `org.apache.kafka.server.streams`, not `kafka.server`, so the link cannot resolve. Use the fully qualified class name (as the Scala test's class-level documentation already does).
##########
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:
This assertion does not prove that the transient failure armed or extended the back-off. The heartbeat that solicited the push already arms the same per-group window, so an immediate heartbeat is suppressed even if transient-failure handling does nothing. Arrange for the solicitation window to expire (or control the broker clock/back-off), then verify suppression after the failure and eventual re-solicitation; that also distinguishes transient handling from a ratcheted failed 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: jira-unsubscribe@kafka.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] MINOR: Add broker integration tests for topology-description plugin failure surfacing [kafka]
Posted by "lucasbru (via GitHub)" <gi...@apache.org>.
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: jira-unsubscribe@kafka.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] MINOR: Add broker integration tests for topology-description plugin failure surfacing [kafka]
Posted by "lucasbru (via GitHub)" <gi...@apache.org>.
lucasbru commented on code in PR #23670:
URL: https://github.com/apache/kafka/pull/23670#discussion_r4165446247
##########
core/src/test/scala/unit/kafka/server/StreamsGroupTopologyDescriptionRequestTest.scala:
##########
@@ -321,6 +323,282 @@ 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()
+ }
+ }
+
+ @Timeout(180)
+ @ClusterTest(serverProperties = Array(
+ new ClusterConfigProperty(
+ key = GroupCoordinatorConfig.STREAMS_GROUP_TOPOLOGY_DESCRIPTION_PLUGIN_CLASS_CONFIG,
+ value = "kafka.server.FailingTopologyDescriptionPlugin")
+ ))
Review Comment:
Fair catch - raised group.streams.session.timeout.ms (and its max) to 120s for this test, well above the 40s sleep.
--
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-unsubscribe@kafka.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] MINOR: Add broker integration tests for topology-description plugin failure surfacing [kafka]
Posted by "lucasbru (via GitHub)" <gi...@apache.org>.
lucasbru merged PR #23670:
URL: https://github.com/apache/kafka/pull/23670
--
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-unsubscribe@kafka.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org