Examples with TIMED_OUT com.netflix.conductor.common.metadata.tasks.Task.Status.TIMED_OUT used on opensource projects

Example 1 with TIMED_OUT

use of com.netflix.conductor.common.metadata.tasks.Task.Status.TIMED_OUT in project conductor by Netflix.

the class DeciderService method decide.

private DeciderOutcome decide(final Workflow workflow, List<Task> preScheduledTasks) throws TerminateWorkflowException {
    DeciderOutcome outcome = new DeciderOutcome();
    if (workflow.getStatus().isTerminal()) {
        // you cannot evaluate a terminal workflow
        LOGGER.debug("Workflow {} is already finished. Reason: {}", workflow, workflow.getReasonForIncompletion());
        return outcome;
    }
    checkWorkflowTimeout(workflow);
    if (workflow.getStatus().equals(WorkflowStatus.PAUSED)) {
        LOGGER.debug("Workflow " + workflow.getWorkflowId() + " is paused");
        return outcome;
    }
    // Filter the list of tasks and include only tasks that are not retried, not executed
    // marked to be skipped and not part of System tasks that is DECISION, FORK, JOIN
    // This list will be empty for a new workflow being started
    List<Task> pendingTasks = workflow.getTasks().stream().filter(isNonPendingTask).collect(Collectors.toList());
    // Get all the tasks that have not completed their lifecycle yet
    // This list will be empty for a new workflow
    Set<String> executedTaskRefNames = workflow.getTasks().stream().filter(Task::isExecuted).map(Task::getReferenceTaskName).collect(Collectors.toSet());
    Map<String, Task> tasksToBeScheduled = new LinkedHashMap<>();
    preScheduledTasks.forEach(preScheduledTask -> {
        tasksToBeScheduled.put(preScheduledTask.getReferenceTaskName(), preScheduledTask);
    });
    // A new workflow does not enter this code branch
    for (Task pendingTask : pendingTasks) {
        if (SystemTaskType.is(pendingTask.getTaskType()) && !pendingTask.getStatus().isTerminal()) {
            tasksToBeScheduled.putIfAbsent(pendingTask.getReferenceTaskName(), pendingTask);
            executedTaskRefNames.remove(pendingTask.getReferenceTaskName());
        }
        Optional<TaskDef> taskDefinition = pendingTask.getTaskDefinition();
        if (!taskDefinition.isPresent()) {
            taskDefinition = Optional.ofNullable(workflow.getWorkflowDefinition().getTaskByRefName(pendingTask.getReferenceTaskName())).map(WorkflowTask::getTaskDefinition);
        }
        if (taskDefinition.isPresent()) {
            checkTaskTimeout(taskDefinition.get(), pendingTask);
            checkTaskPollTimeout(taskDefinition.get(), pendingTask);
            // If the task has not been updated for "responseTimeoutSeconds" then mark task as TIMED_OUT
            if (isResponseTimedOut(taskDefinition.get(), pendingTask)) {
                timeoutTask(taskDefinition.get(), pendingTask);
            }
        }
        if (!pendingTask.getStatus().isSuccessful()) {
            WorkflowTask workflowTask = pendingTask.getWorkflowTask();
            if (workflowTask == null) {
                workflowTask = workflow.getWorkflowDefinition().getTaskByRefName(pendingTask.getReferenceTaskName());
            }
            Optional<Task> retryTask = retry(taskDefinition.orElse(null), workflowTask, pendingTask, workflow);
            if (retryTask.isPresent()) {
                tasksToBeScheduled.put(retryTask.get().getReferenceTaskName(), retryTask.get());
                executedTaskRefNames.remove(retryTask.get().getReferenceTaskName());
                outcome.tasksToBeUpdated.add(pendingTask);
            } else {
                pendingTask.setStatus(COMPLETED_WITH_ERRORS);
            }
        }
        if (!pendingTask.isExecuted() && !pendingTask.isRetried() && pendingTask.getStatus().isTerminal()) {
            pendingTask.setExecuted(true);
            List<Task> nextTasks = getNextTask(workflow, pendingTask);
            if (pendingTask.isLoopOverTask() && !TaskType.DO_WHILE.name().equals(pendingTask.getTaskType()) && !nextTasks.isEmpty()) {
                nextTasks = filterNextLoopOverTasks(nextTasks, pendingTask, workflow);
            }
            nextTasks.forEach(nextTask -> tasksToBeScheduled.putIfAbsent(nextTask.getReferenceTaskName(), nextTask));
            outcome.tasksToBeUpdated.add(pendingTask);
            LOGGER.debug("Scheduling Tasks from {}, next = {} for workflowId: {}", pendingTask.getTaskDefName(), nextTasks.stream().map(Task::getTaskDefName).collect(Collectors.toList()), workflow.getWorkflowId());
        }
    }
    // All the tasks that need to scheduled are added to the outcome, in case of
    List<Task> unScheduledTasks = tasksToBeScheduled.values().stream().filter(task -> !executedTaskRefNames.contains(task.getReferenceTaskName())).collect(Collectors.toList());
    if (!unScheduledTasks.isEmpty()) {
        LOGGER.debug("Scheduling Tasks: {} for workflow: {}", unScheduledTasks.stream().map(Task::getTaskDefName).collect(Collectors.toList()), workflow.getWorkflowId());
        outcome.tasksToBeScheduled.addAll(unScheduledTasks);
    }
    if (containsSuccessfulTerminateTask.test(workflow) || (outcome.tasksToBeScheduled.isEmpty() && checkForWorkflowCompletion(workflow))) {
        LOGGER.debug("Marking workflow: {} as complete.", workflow);
        outcome.isComplete = true;
    }
    return outcome;
}

Example 2 with TIMED_OUT

use of com.netflix.conductor.common.metadata.tasks.Task.Status.TIMED_OUT in project conductor by Netflix.

the class AbstractWorkflowServiceTest method testSimpleWorkflowWithResponseTimeout.

@Test
public void testSimpleWorkflowWithResponseTimeout() throws Exception {
    createWFWithResponseTimeout();
    String correlationId = "unit_test_1";
    Map<String, Object> workflowInput = new HashMap<>();
    String inputParam1 = "p1 value";
    workflowInput.put("param1", inputParam1);
    workflowInput.put("param2", "p2 value");
    String workflowId = startOrLoadWorkflowExecution("RTOWF", 1, correlationId, workflowInput, null, null);
    logger.debug("testSimpleWorkflowWithResponseTimeout.wfid=" + workflowId);
    assertNotNull(workflowId);
    Workflow workflow = workflowExecutionService.getExecutionStatus(workflowId, true);
    assertNotNull(workflow);
    assertEquals(RUNNING, workflow.getStatus());
    // The very first task is the one that should be scheduled.
    assertEquals(1, workflow.getTasks().size());
    assertEquals(1, queueDAO.getSize("task_rt"));
    // Polling for the first task should return the first task
    Task task = workflowExecutionService.poll("task_rt", "task1.junit.worker.testTimeout");
    assertNotNull(task);
    assertEquals("task_rt", task.getTaskType());
    assertTrue(workflowExecutionService.ackTaskReceived(task.getTaskId()));
    assertEquals(workflowId, task.getWorkflowInstanceId());
    // As the task_rt is out of the queue, the next poll should not get it
    Task nullTask = workflowExecutionService.poll("task_rt", "task1.junit.worker.testTimeout");
    assertNull(nullTask);
    Thread.sleep(10000);
    workflowExecutor.decide(workflowId);
    assertEquals(1, queueDAO.getSize("task_rt"));
    // The first task would be timed_out and a new task will be scheduled
    workflow = workflowExecutionService.getExecutionStatus(workflowId, true);
    assertNotNull(workflow);
    assertEquals(RUNNING, workflow.getStatus());
    assertEquals(2, workflow.getTasks().size());
    assertTrue(workflow.getTasks().stream().allMatch(t -> t.getReferenceTaskName().equals("task_rt_t1")));
    assertEquals(TIMED_OUT, workflow.getTasks().get(0).getStatus());
    assertEquals(SCHEDULED, workflow.getTasks().get(1).getStatus());
    // Polling now should get the same task back because it should have been put back in the queue
    Task taskAgain = workflowExecutionService.poll("task_rt", "task1.junit.worker");
    assertNotNull(taskAgain);
    // update task with callback after seconds greater than the response timeout
    taskAgain.setStatus(IN_PROGRESS);
    taskAgain.setCallbackAfterSeconds(2);
    workflowExecutionService.updateTask(taskAgain);
    workflow = workflowExecutionService.getExecutionStatus(workflowId, true);
    assertNotNull(workflow);
    assertEquals(WorkflowStatus.RUNNING, workflow.getStatus());
    assertEquals(2, workflow.getTasks().size());
    assertEquals(SCHEDULED, workflow.getTasks().get(1).getStatus());
    // wait for callback after seconds which is longer than response timeout seconds and then call decide
    Thread.sleep(2010);
    // Ensure unacks are processed.
    queueDAO.processUnacks(taskAgain.getTaskDefName());
    workflowExecutor.decide(workflowId);
    workflow = workflowExecutionService.getExecutionStatus(workflowId, true);
    assertNotNull(workflow);
    // Poll for task again
    taskAgain = workflowExecutionService.poll("task_rt", "task1.junit.worker");
    assertNotNull(taskAgain);
    taskAgain.getOutputData().put("op", "task1.Done");
    taskAgain.setStatus(COMPLETED);
    workflowExecutionService.updateTask(taskAgain);
    // poll for next task
    task = workflowExecutionService.poll("junit_task_2", "task2.junit.worker.testTimeout");
    assertNotNull(task);
    assertEquals("junit_task_2", task.getTaskType());
    assertTrue(workflowExecutionService.ackTaskReceived(task.getTaskId()));
    task.setStatus(COMPLETED);
    task.setReasonForIncompletion("unit test failure");
    workflowExecutionService.updateTask(task);
    workflow = workflowExecutionService.getExecutionStatus(workflowId, true);
    assertNotNull(workflow);
    assertEquals(WorkflowStatus.COMPLETED, workflow.getStatus());
}

Example 3 with TIMED_OUT

use of com.netflix.conductor.common.metadata.tasks.Task.Status.TIMED_OUT in project conductor by Netflix.

the class WorkflowExecutor method retry.

/**
 * Gets the last instance of each failed task and reschedule each
 * Gets all cancelled tasks and schedule all of them except JOIN (join should change status to INPROGRESS)
 * Switch workflow back to RUNNING status and call decider.
 *
 * @param workflowId the id of the workflow to be retried
 */
public void retry(String workflowId, boolean resumeSubworkflowTasks) {
    Workflow workflow = executionDAOFacade.getWorkflowById(workflowId, true);
    if (!workflow.getStatus().isTerminal()) {
        throw new ApplicationException(CONFLICT, "Workflow is still running.  status=" + workflow.getStatus());
    }
    if (workflow.getTasks().isEmpty()) {
        throw new ApplicationException(CONFLICT, "Workflow has not started yet");
    }
    if (resumeSubworkflowTasks) {
        Optional<Task> lTask = workflow.getTasks().stream().filter(this::findLastFailedOrTimeOutTask).findFirst();
        if (lTask.isPresent()) {
            workflow = findLastFailedSubWorkflow(lTask.get(), workflow);
        }
    }
    // Get all FAILED or CANCELED tasks that are not COMPLETED (or reach other terminal states) on further executions.
    // // Eg: for Seq of tasks task1.CANCELED, task1.COMPLETED, task1 shouldn't be retried.
    // Throw an exception if there are no FAILED tasks.
    // Handle JOIN task CANCELED status as special case.
    Map<String, Task> retriableMap = new HashMap<>();
    for (Task task : workflow.getTasks()) {
        switch(task.getStatus()) {
            case FAILED:
            case FAILED_WITH_TERMINAL_ERROR:
            case TIMED_OUT:
                retriableMap.put(task.getReferenceTaskName(), task);
                break;
            case CANCELED:
                if (task.getTaskType().equalsIgnoreCase(TaskType.JOIN.toString()) || task.getTaskType().equalsIgnoreCase(TaskType.DO_WHILE.toString())) {
                    task.setStatus(IN_PROGRESS);
                // Task doesn't have to be updated yet. Will be updated along with other Workflow tasks downstream.
                } else {
                    retriableMap.put(task.getReferenceTaskName(), task);
                }
                break;
            default:
                retriableMap.remove(task.getReferenceTaskName());
                break;
        }
    }
    if (retriableMap.values().size() == 0) {
        throw new ApplicationException(CONFLICT, "There are no retriable tasks! Use restart if you want to attempt entire workflow execution again.");
    }
    // Update Workflow with new status.
    // This should load Workflow from archive, if archived.
    workflow.setStatus(WorkflowStatus.RUNNING);
    workflow.setLastRetriedTime(System.currentTimeMillis());
    // Add to decider queue
    queueDAO.push(DECIDER_QUEUE, workflow.getWorkflowId(), workflow.getPriority(), config.getSweepFrequency());
    executionDAOFacade.updateWorkflow(workflow);
    // taskToBeRescheduled would set task `retried` to true, and hence it's important to updateTasks after obtaining task copy from taskToBeRescheduled.
    final Workflow finalWorkflow = workflow;
    List<Task> retriableTasks = retriableMap.values().stream().sorted(Comparator.comparingInt(Task::getSeq)).map(task -> taskToBeRescheduled(finalWorkflow, task)).collect(Collectors.toList());
    dedupAndAddTasks(workflow, retriableTasks);
    // Note: updateTasks before updateWorkflow might fail when Workflow is archived and doesn't exist in primary store.
    executionDAOFacade.updateTasks(workflow.getTasks());
    scheduleTask(workflow, retriableTasks);
    decide(workflowId);
    if (resumeSubworkflowTasks) {
        updateParentWorkflowRecursively(workflow);
    } else if (StringUtils.isNotEmpty(workflow.getParentWorkflowId())) {
        updateParentWorkflow(workflow);
        decide(workflow.getParentWorkflowId());
    }
}