You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@kafka.apache.org by da...@apache.org on 2024/08/09 08:06:45 UTC

(kafka) branch trunk updated: KAFKA-17267; Don't return REQUEST_TIMED_OUT for OFFSET_FETCHes (#16825)

This is an automated email from the ASF dual-hosted git repository.

dajac pushed a commit to branch trunk
in repository https://gitbox.apache.org/repos/asf/kafka.git


The following commit(s) were added to refs/heads/trunk by this push:
     new 7a8edffad10 KAFKA-17267; Don't return REQUEST_TIMED_OUT for OFFSET_FETCHes (#16825)
7a8edffad10 is described below

commit 7a8edffad10c5b58fee19985f157942425b103a9
Author: Sean Quah <sq...@confluent.io>
AuthorDate: Fri Aug 9 09:06:38 2024 +0100

    KAFKA-17267; Don't return REQUEST_TIMED_OUT for OFFSET_FETCHes (#16825)
    
    When handling an OFFSET_FETCH request requiring stable offsets, the new
    group coordinator may encounter a timeout under some circumstances, such
    as a zombie coordinator or a lagging __consumer_offsets replica that has
    not yet dropped out of the ISR. Existing and older clients do not expect
    the REQUEST_TIMED_OUT error code won't retry, so remap it to
    NOT_COORDINATOR to trigger a coordinator lookup and retry.
    
    Reviewers: David Jacot <dj...@confluent.io>
---
 .../coordinator/group/GroupCoordinatorService.java | 57 ++++++++++++-
 .../group/GroupCoordinatorServiceTest.java         | 97 ++++++++++++++++++++++
 2 files changed, 152 insertions(+), 2 deletions(-)

diff --git a/group-coordinator/src/main/java/org/apache/kafka/coordinator/group/GroupCoordinatorService.java b/group-coordinator/src/main/java/org/apache/kafka/coordinator/group/GroupCoordinatorService.java
index b7f44544b30..9316ad3837f 100644
--- a/group-coordinator/src/main/java/org/apache/kafka/coordinator/group/GroupCoordinatorService.java
+++ b/group-coordinator/src/main/java/org/apache/kafka/coordinator/group/GroupCoordinatorService.java
@@ -825,7 +825,11 @@ public class GroupCoordinatorService implements GroupCoordinator {
                     Collections.emptyList(),
                     coordinator.fetchOffsets(request, Long.MAX_VALUE)
                 )
-            );
+            ).exceptionally(exception -> handleOffsetFetchException(
+                "fetch-offsets",
+                request,
+                exception
+            ));
         } else {
             return runtime.scheduleReadOperation(
                 "fetch-offsets",
@@ -876,7 +880,11 @@ public class GroupCoordinatorService implements GroupCoordinator {
                     Collections.emptyList(),
                     coordinator.fetchAllOffsets(request, Long.MAX_VALUE)
                 )
-            );
+            ).exceptionally(exception -> handleOffsetFetchException(
+                "fetch-all-offsets",
+                request,
+                exception
+            ));
         } else {
             return runtime.scheduleReadOperation(
                 "fetch-all-offsets",
@@ -1217,4 +1225,49 @@ public class GroupCoordinatorService implements GroupCoordinator {
                 return handler.apply(apiError.error(), apiError.message());
         }
     }
+
+    /**
+     * This is the handler used by offset fetch operations to convert errors to coordinator errors.
+     * The handler also handles and log unexpected errors.
+     *
+     * @param operationName     The name of the operation.
+     * @param request           The OffsetFetchRequestGroup request.
+     * @param exception         The exception to handle.
+     * @return The OffsetFetchRequestGroup response.
+     */
+    private OffsetFetchResponseData.OffsetFetchResponseGroup handleOffsetFetchException(
+        String operationName,
+        OffsetFetchRequestData.OffsetFetchRequestGroup request,
+        Throwable exception
+    ) {
+        ApiError apiError = ApiError.fromThrowable(exception);
+
+        switch (apiError.error()) {
+            case UNKNOWN_TOPIC_OR_PARTITION:
+            case NOT_ENOUGH_REPLICAS:
+            case REQUEST_TIMED_OUT:
+                // Remap REQUEST_TIMED_OUT to NOT_COORDINATOR, since consumers on versions prior
+                // to 3.9 do not expect the error and won't retry the request. NOT_COORDINATOR
+                // additionally triggers coordinator re-lookup, which is necessary if the client is
+                // talking to a zombie coordinator.
+                //
+                // While handleOperationException does remap UNKNOWN_TOPIC_OR_PARTITION,
+                // NOT_ENOUGH_REPLICAS and REQUEST_TIMED_OUT to COORDINATOR_NOT_AVAILABLE,
+                // COORDINATOR_NOT_AVAILABLE is also not handled by consumers on versions prior to
+                // 3.9.
+                return new OffsetFetchResponseData.OffsetFetchResponseGroup()
+                    .setGroupId(request.groupId())
+                    .setErrorCode(Errors.NOT_COORDINATOR.code());
+
+            default:
+                return handleOperationException(
+                    operationName,
+                    request,
+                    exception,
+                    (error, __) -> new OffsetFetchResponseData.OffsetFetchResponseGroup()
+                        .setGroupId(request.groupId())
+                        .setErrorCode(error.code())
+                );
+        }
+    }
 }
diff --git a/group-coordinator/src/test/java/org/apache/kafka/coordinator/group/GroupCoordinatorServiceTest.java b/group-coordinator/src/test/java/org/apache/kafka/coordinator/group/GroupCoordinatorServiceTest.java
index f3f9aa6c969..4c7b6e704f0 100644
--- a/group-coordinator/src/test/java/org/apache/kafka/coordinator/group/GroupCoordinatorServiceTest.java
+++ b/group-coordinator/src/test/java/org/apache/kafka/coordinator/group/GroupCoordinatorServiceTest.java
@@ -1162,6 +1162,56 @@ public class GroupCoordinatorServiceTest {
         );
     }
 
+    @ParameterizedTest
+    @CsvSource({
+        "UNKNOWN_TOPIC_OR_PARTITION, NOT_COORDINATOR",
+        "NOT_ENOUGH_REPLICAS, NOT_COORDINATOR",
+        "REQUEST_TIMED_OUT, NOT_COORDINATOR",
+        "NOT_LEADER_OR_FOLLOWER, NOT_COORDINATOR",
+        "KAFKA_STORAGE_ERROR, NOT_COORDINATOR",
+    })
+    public void testFetchOffsetsWithWrappedError(
+        Errors error,
+        Errors expectedError
+    ) throws ExecutionException, InterruptedException {
+        CoordinatorRuntime<GroupCoordinatorShard, CoordinatorRecord> runtime = mockRuntime();
+        GroupCoordinatorService service = new GroupCoordinatorService(
+            new LogContext(),
+            createConfig(),
+            runtime,
+            new GroupCoordinatorMetrics()
+        );
+
+        service.startup(() -> 1);
+
+        OffsetFetchRequestData.OffsetFetchRequestGroup request =
+            new OffsetFetchRequestData.OffsetFetchRequestGroup()
+                .setGroupId("group")
+                .setTopics(Collections.singletonList(new OffsetFetchRequestData.OffsetFetchRequestTopics()
+                    .setName("foo")
+                    .setPartitionIndexes(Collections.singletonList(0))));
+
+        when(runtime.scheduleWriteOperation(
+            ArgumentMatchers.eq("fetch-offsets"),
+            ArgumentMatchers.eq(new TopicPartition("__consumer_offsets", 0)),
+            ArgumentMatchers.eq(Duration.ofMillis(5000)),
+            ArgumentMatchers.any()
+        )).thenReturn(FutureUtils.failedFuture(new CompletionException(error.exception())));
+
+        CompletableFuture<OffsetFetchResponseData.OffsetFetchResponseGroup> future = service.fetchOffsets(
+            requestContext(ApiKeys.OFFSET_FETCH),
+            request,
+            true
+        );
+
+        assertEquals(
+            new OffsetFetchResponseData.OffsetFetchResponseGroup()
+                .setGroupId("group")
+                .setErrorCode(expectedError.code()),
+            future.get()
+        );
+    }
+
     @ParameterizedTest
     @ValueSource(booleans = {true, false})
     public void testFetchAllOffsets(
@@ -1245,6 +1295,53 @@ public class GroupCoordinatorServiceTest {
         );
     }
 
+    @ParameterizedTest
+    @CsvSource({
+        "UNKNOWN_TOPIC_OR_PARTITION, NOT_COORDINATOR",
+        "NOT_ENOUGH_REPLICAS, NOT_COORDINATOR",
+        "REQUEST_TIMED_OUT, NOT_COORDINATOR",
+        "NOT_LEADER_OR_FOLLOWER, NOT_COORDINATOR",
+        "KAFKA_STORAGE_ERROR, NOT_COORDINATOR",
+    })
+    public void testFetchAllOffsetsWithWrappedError(
+        Errors error,
+        Errors expectedError
+    ) throws ExecutionException, InterruptedException {
+        CoordinatorRuntime<GroupCoordinatorShard, CoordinatorRecord> runtime = mockRuntime();
+        GroupCoordinatorService service = new GroupCoordinatorService(
+            new LogContext(),
+            createConfig(),
+            runtime,
+            new GroupCoordinatorMetrics()
+        );
+
+        service.startup(() -> 1);
+
+        OffsetFetchRequestData.OffsetFetchRequestGroup request =
+            new OffsetFetchRequestData.OffsetFetchRequestGroup()
+                .setGroupId("group");
+
+        when(runtime.scheduleWriteOperation(
+            ArgumentMatchers.eq("fetch-all-offsets"),
+            ArgumentMatchers.eq(new TopicPartition("__consumer_offsets", 0)),
+            ArgumentMatchers.eq(Duration.ofMillis(5000)),
+            ArgumentMatchers.any()
+            )).thenReturn(FutureUtils.failedFuture(new CompletionException(error.exception())));
+
+        CompletableFuture<OffsetFetchResponseData.OffsetFetchResponseGroup> future = service.fetchAllOffsets(
+            requestContext(ApiKeys.OFFSET_FETCH),
+            request,
+            true
+        );
+
+        assertEquals(
+            new OffsetFetchResponseData.OffsetFetchResponseGroup()
+                .setGroupId("group")
+                .setErrorCode(expectedError.code()),
+            future.get()
+        );
+    }
+
     @Test
     public void testLeaveGroup() throws Exception {
         CoordinatorRuntime<GroupCoordinatorShard, CoordinatorRecord> runtime = mockRuntime();