Github user StephanEwen commented on a diff in the pull request: https://github.com/apache/flink/pull/2652#discussion_r83842005 --- Diff: flink-runtime/src/test/java/org/apache/flink/runtime/taskmanager/TaskTest.java --- @@ -565,6 +568,128 @@ public void testOnPartitionStateUpdate() throws Exception { verify(inputGate, times(1)).retriggerPartitionRequest(eq(partitionId.getPartitionId())); } + /** + * Tests that interrupt happens via watch dog if canceller is stuck in cancel. + * Task cancellation blocks the task canceller. Interrupt after cancel via + * cancellation watch dog. + */ + @Test + public void testWatchDogInterruptsTask() throws Exception { + Configuration config = new Configuration(); + config.setLong(TaskOptions.CANCELLATION_INTERVAL.key(), 5); + config.setLong(TaskOptions.CANCELLATION_TIMEOUT.key(), 50); + + Task task = createTask(InvokableBlockingInCancel.class, config); + task.startTaskThread(); + + awaitLatch.await(); + + task.cancelExecution(); + + triggerLatch.await(); + + // No fatal error + for (Object msg : taskManagerMessages) { + assertEquals(false, msg instanceof TaskManagerMessages.FatalError); + } + } + + /** + * The invoke() method holds a lock (trigger awaitLatch after acquisition) + * and cancel cannot complete because it also tries to acquire the same lock. + * This is resolved by the watch dog, no fatal error. + */ + @Test + public void testInterruptableSharedLockInInvokeAndCancel() throws Exception { + Configuration config = new Configuration(); + config.setLong(TaskOptions.CANCELLATION_INTERVAL.key(), 5); + config.setLong(TaskOptions.CANCELLATION_TIMEOUT.key(), 50); + + Task task = createTask(InvokableInterruptableSharedLockInInvokeAndCancel.class, config); + task.startTaskThread(); + + awaitLatch.await(); + + task.cancelExecution(); + + triggerLatch.await(); + + // No fatal error + for (Object msg : taskManagerMessages) { + assertEquals(false, msg instanceof TaskManagerMessages.FatalError); --- End diff -- You can also give a message to `assertFalse` - I like assertEquals for printing the expected value, but if the expected value is false, the former seems more natural to me...
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---