mjsax commented on a change in pull request #8589: URL: https://github.com/apache/kafka/pull/8589#discussion_r430826643
########## File path: core/src/main/scala/kafka/tools/StreamsResetter.java ########## @@ -186,9 +192,19 @@ private void validateNoActiveConsumers(final String groupId, final List<MemberDescription> members = new ArrayList<>(describeResult.describedGroups().get(groupId).get().members()); if (!members.isEmpty()) { - throw new IllegalStateException("Consumer group '" + groupId + "' is still active " - + "and has following members: " + members + ". " - + "Make sure to stop all running application instances before running the reset tool."); + if (options.has(forceOption)) { + System.out.println("Force deleting all active members in the group: " + groupId); + try { + adminClient.removeMembersFromConsumerGroup(groupId, new RemoveMembersFromConsumerGroupOptions()).all().get(); + } catch (Exception e) { + throw e; + } + } else { + throw new IllegalStateException("Consumer group '" + groupId + "' is still active " + + "and has following members: " + members + ". " + + "Make sure to stop all running application instances before running the reset tool." + + "Try set '--force' in the cmdline to force delete active members."); Review comment: ```suggestion "You can use option '--force' to remove active members from the group."); ``` ########## File path: core/src/main/scala/kafka/tools/StreamsResetter.java ########## @@ -186,9 +192,19 @@ private void validateNoActiveConsumers(final String groupId, final List<MemberDescription> members = new ArrayList<>(describeResult.describedGroups().get(groupId).get().members()); if (!members.isEmpty()) { - throw new IllegalStateException("Consumer group '" + groupId + "' is still active " - + "and has following members: " + members + ". " - + "Make sure to stop all running application instances before running the reset tool."); + if (options.has(forceOption)) { + System.out.println("Force deleting all active members in the group: " + groupId); + try { + adminClient.removeMembersFromConsumerGroup(groupId, new RemoveMembersFromConsumerGroupOptions()).all().get(); + } catch (Exception e) { + throw e; + } + } else { + throw new IllegalStateException("Consumer group '" + groupId + "' is still active " + + "and has following members: " + members + ". " + + "Make sure to stop all running application instances before running the reset tool." + + "Try set '--force' in the cmdline to force delete active members."); Review comment: ```suggestion + " You can use option '--force' to remove active members from the group."); ``` ########## File path: core/src/main/scala/kafka/tools/StreamsResetter.java ########## @@ -186,9 +192,19 @@ private void validateNoActiveConsumers(final String groupId, final List<MemberDescription> members = new ArrayList<>(describeResult.describedGroups().get(groupId).get().members()); if (!members.isEmpty()) { - throw new IllegalStateException("Consumer group '" + groupId + "' is still active " - + "and has following members: " + members + ". " - + "Make sure to stop all running application instances before running the reset tool."); + if (options.has(forceOption)) { + System.out.println("Force deleting all active members in the group: " + groupId); + try { + adminClient.removeMembersFromConsumerGroup(groupId, new RemoveMembersFromConsumerGroupOptions()).all().get(); + } catch (Exception e) { + throw e; + } + } else { + throw new IllegalStateException("Consumer group '" + groupId + "' is still active " + + "and has following members: " + members + ". " + + "Make sure to stop all running application instances before running the reset tool." + Review comment: ```suggestion + "Make sure to stop all running application instances before running the reset tool." ``` ########## File path: core/src/main/scala/kafka/tools/StreamsResetter.java ########## @@ -236,6 +252,8 @@ private void parseArguments(final String[] args) { .withRequiredArg() .ofType(String.class) .describedAs("file name"); + forceOption = optionParser.accepts("force", "Force remove members when long session time out has been configured, " + Review comment: ```suggestion forceOption = optionParser.accepts("force", "Force the removal of members of the consumer group. " + ``` ########## File path: core/src/main/scala/kafka/tools/StreamsResetter.java ########## @@ -236,6 +252,8 @@ private void parseArguments(final String[] args) { .withRequiredArg() .ofType(String.class) .describedAs("file name"); + forceOption = optionParser.accepts("force", "Force remove members when long session time out has been configured, " + + "please make sure to shut down all stream applications when this option is specified to avoid unexpected rebalances."); Review comment: ```suggestion "Make sure to shut down all stream applications when this option is specified to avoid unexpected rebalances."); ``` ########## File path: core/src/main/scala/kafka/tools/StreamsResetter.java ########## @@ -236,6 +252,8 @@ private void parseArguments(final String[] args) { .withRequiredArg() .ofType(String.class) .describedAs("file name"); + forceOption = optionParser.accepts("force", "Force remove members when long session time out has been configured, " + Review comment: ```suggestion forceOption = optionParser.accepts("force", "Force the removal of members of the consumer group (intended to remove stopped members if a long session timeout was used). " + ``` ########## File path: streams/src/test/java/org/apache/kafka/streams/integration/AbstractResetIntegrationTest.java ########## @@ -261,6 +261,42 @@ public void shouldNotAllowToResetWhenIntermediateTopicAbsent() throws Exception Assert.assertEquals(1, exitCode); } + public void testResetWhenLongSessionTimeoutConfiguredWithForceOption() throws Exception { + appID = testId + "-with-force-option"; + streamsConfig.put(StreamsConfig.APPLICATION_ID_CONFIG, appID); + streamsConfig.put(ConsumerConfig.SESSION_TIMEOUT_MS_CONFIG, "" + STREAMS_CONSUMER_TIMEOUT * 100); + + // Run + streams = new KafkaStreams(setupTopologyWithoutIntermediateUserTopic(), streamsConfig); + streams.start(); + final List<KeyValue<Long, Long>> result = IntegrationTestUtils.waitUntilMinKeyValueRecordsReceived(resultConsumerConfig, OUTPUT_TOPIC, 10); + + streams.close(); + + // RESET + streams = new KafkaStreams(setupTopologyWithoutIntermediateUserTopic(), streamsConfig); + streams.cleanUp(); + + // Reset would fail since long session timeout has been configured + final boolean cleanResult = tryCleanGlobal(false, null, null); + Assert.assertEquals(false, cleanResult); + + // Reset will success with --force, it will force delete active members on broker side + cleanGlobal(false, "--force", null); + + waitForEmptyConsumerGroup(adminClient, appID, TIMEOUT_MULTIPLIER * CLEANUP_CONSUMER_TIMEOUT); + + assertInternalTopicsGotDeleted(null); + + // RE-RUN Review comment: Why do we need this part? Seems sufficient to end the test here? ########## File path: streams/src/test/java/org/apache/kafka/streams/integration/AbstractResetIntegrationTest.java ########## @@ -261,6 +261,42 @@ public void shouldNotAllowToResetWhenIntermediateTopicAbsent() throws Exception Assert.assertEquals(1, exitCode); } + public void testResetWhenLongSessionTimeoutConfiguredWithForceOption() throws Exception { + appID = testId + "-with-force-option"; + streamsConfig.put(StreamsConfig.APPLICATION_ID_CONFIG, appID); + streamsConfig.put(ConsumerConfig.SESSION_TIMEOUT_MS_CONFIG, "" + STREAMS_CONSUMER_TIMEOUT * 100); + + // Run + streams = new KafkaStreams(setupTopologyWithoutIntermediateUserTopic(), streamsConfig); + streams.start(); + final List<KeyValue<Long, Long>> result = IntegrationTestUtils.waitUntilMinKeyValueRecordsReceived(resultConsumerConfig, OUTPUT_TOPIC, 10); + + streams.close(); + + // RESET + streams = new KafkaStreams(setupTopologyWithoutIntermediateUserTopic(), streamsConfig); + streams.cleanUp(); + + // Reset would fail since long session timeout has been configured + final boolean cleanResult = tryCleanGlobal(false, null, null); + Assert.assertEquals(false, cleanResult); + + // Reset will success with --force, it will force delete active members on broker side + cleanGlobal(false, "--force", null); + + waitForEmptyConsumerGroup(adminClient, appID, TIMEOUT_MULTIPLIER * CLEANUP_CONSUMER_TIMEOUT); Review comment: With `cleanGlobal` and `--force` the consumer group could be empty when `cleanGlobal` returns, right? Hence, we should do this assertion without timeout or retries? ########## File path: clients/src/main/java/org/apache/kafka/clients/admin/KafkaAdminClient.java ########## @@ -3623,22 +3641,26 @@ public RemoveMembersFromConsumerGroupResult removeMembersFromConsumerGroup(Strin ConsumerGroupOperationContext<Map<MemberIdentity, Errors>, RemoveMembersFromConsumerGroupOptions> context = new ConsumerGroupOperationContext<>(groupId, options, deadline, future); - Call findCoordinatorCall = getFindCoordinatorCall(context, - () -> getRemoveMembersFromGroupCall(context)); + List<MemberIdentity> members; + if (options.removeAll()) { + members = getMembersFromGroup(groupId); + } else { + members = options.members().stream().map(MemberToRemove::toMemberIdentity).collect(Collectors.toList()); + } + Call findCoordinatorCall = getFindCoordinatorCall(context, () -> getRemoveMembersFromGroupCall(context, members)); runnable.call(findCoordinatorCall, startFindCoordinatorMs); return new RemoveMembersFromConsumerGroupResult(future, options.members()); Review comment: If `option.members()` is empty, it implies that we do a `removeAll()` -- hence, should we pass in `members` into the `RemoveMembersFromConsumerGroupResult` instead of `options.members()` ? ########## File path: clients/src/main/java/org/apache/kafka/clients/admin/KafkaAdminClient.java ########## @@ -3623,22 +3641,26 @@ public RemoveMembersFromConsumerGroupResult removeMembersFromConsumerGroup(Strin ConsumerGroupOperationContext<Map<MemberIdentity, Errors>, RemoveMembersFromConsumerGroupOptions> context = new ConsumerGroupOperationContext<>(groupId, options, deadline, future); - Call findCoordinatorCall = getFindCoordinatorCall(context, - () -> getRemoveMembersFromGroupCall(context)); + List<MemberIdentity> members; + if (options.removeAll()) { + members = getMembersFromGroup(groupId); + } else { + members = options.members().stream().map(MemberToRemove::toMemberIdentity).collect(Collectors.toList()); + } + Call findCoordinatorCall = getFindCoordinatorCall(context, () -> getRemoveMembersFromGroupCall(context, members)); runnable.call(findCoordinatorCall, startFindCoordinatorMs); return new RemoveMembersFromConsumerGroupResult(future, options.members()); } - private Call getRemoveMembersFromGroupCall(ConsumerGroupOperationContext<Map<MemberIdentity, Errors>, RemoveMembersFromConsumerGroupOptions> context) { + private Call getRemoveMembersFromGroupCall(ConsumerGroupOperationContext<Map<MemberIdentity, Errors>, + RemoveMembersFromConsumerGroupOptions> context, List<MemberIdentity> members) { Review comment: nit: fix formatting: ``` private Call getRemoveMembersFromGroupCall(ConsumerGroupOperationContext<Map<MemberIdentity, Errors>, RemoveMembersFromConsumerGroupOptions> context, List<MemberIdentity> members) { ``` ########## File path: clients/src/main/java/org/apache/kafka/clients/admin/KafkaAdminClient.java ########## @@ -3660,7 +3686,7 @@ void handleResponse(AbstractResponse abstractResponse) { // We set member.id to empty here explicitly, so that the lookup will succeed as user doesn't // know the exact member.id. memberErrors.put(new MemberIdentity() - .setMemberId(JoinGroupRequest.UNKNOWN_MEMBER_ID) + .setMemberId(memberResponse.memberId()) Review comment: Not sure if I understand the change. Also not sure if I can follow the comments. Can you elaborate? ########## File path: clients/src/main/java/org/apache/kafka/clients/admin/RemoveMembersFromConsumerGroupOptions.java ########## @@ -37,7 +38,15 @@ public RemoveMembersFromConsumerGroupOptions(Collection<MemberToRemove> members) this.members = new HashSet<>(members); Review comment: As we have different semantics for an empty collection (it was "remove nothing" originally, and we change it to "remove all"), I am wondering if we should do a check if `members` is empty or not and throw an exception if empty? Or at least log a WARNING that empty implies "remove all" now? ########## File path: clients/src/main/java/org/apache/kafka/clients/admin/RemoveMembersFromConsumerGroupResult.java ########## @@ -51,9 +52,21 @@ if (throwable != null) { result.completeExceptionally(throwable); } else { - for (MemberToRemove memberToRemove : memberInfos) { - if (maybeCompleteExceptionally(memberErrors, memberToRemove.toMemberIdentity(), result)) { - return; + if (removeAll()) { Review comment: Not sure why the `removeAll()` case needs to be handled differently? Can you elaboarte? ########## File path: clients/src/main/java/org/apache/kafka/clients/admin/RemoveMembersFromConsumerGroupResult.java ########## @@ -51,9 +52,21 @@ if (throwable != null) { result.completeExceptionally(throwable); } else { - for (MemberToRemove memberToRemove : memberInfos) { - if (maybeCompleteExceptionally(memberErrors, memberToRemove.toMemberIdentity(), result)) { - return; + if (removeAll()) { Review comment: Not sure why the `removeAll()` case needs to be handled differently? Can you elaborate? ########## File path: clients/src/main/java/org/apache/kafka/clients/admin/RemoveMembersFromConsumerGroupResult.java ########## @@ -66,6 +79,9 @@ * Returns the selected member future. */ public KafkaFuture<Void> memberResult(MemberToRemove member) { + if (removeAll()) { + throw new IllegalArgumentException("The method: memberResult is not applicable in 'removeAll' mode"); Review comment: Why that? I understand that we expect that users don't know the memberId if the so a "remove all"; however, I don't see why we need to disallow this call? Can you elaborate? ########## File path: clients/src/test/java/org/apache/kafka/clients/admin/KafkaAdminClientTest.java ########## @@ -379,6 +380,22 @@ private static MetadataResponse prepareMetadataResponse(Cluster cluster, Errors MetadataResponse.AUTHORIZED_OPERATIONS_OMITTED); } + private static DescribeGroupsResponseData prepareDescribeGroupsResponseData(String groupId, List<String> groupInstances, Review comment: Nit: formatting ``` private static DescribeGroupsResponseData prepareDescribeGroupsResponseData(String groupId, List<String> groupInstances, List<TopicPartition> topicPartitions) { ``` ########## File path: core/src/test/scala/integration/kafka/api/PlaintextAdminIntegrationTest.scala ########## @@ -1017,47 +1017,70 @@ class PlaintextAdminIntegrationTest extends BaseAdminIntegrationTest { assertTrue(0 == list1.errors().get().size()) assertTrue(0 == list1.valid().get().size()) val testTopicName = "test_topic" + val testTopicName1 = testTopicName + "1" + val testTopicName2 = testTopicName + "2" val testNumPartitions = 2 - client.createTopics(Collections.singleton( - new NewTopic(testTopicName, testNumPartitions, 1.toShort))).all().get() - waitForTopics(client, List(testTopicName), List()) + + client.createTopics(util.Arrays.asList(new NewTopic(testTopicName, testNumPartitions, 1.toShort), Review comment: nit: formatting: move `new NewTopic(...)` to next line ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org