1、Worker向Master三次汇报

Worker向Master汇报.png

注意 : RUNNING_INFO更新有问题,因为根本就不会更新PID,因为都没有传递过去PID,所以这个有什么用呢?PID还是最后FINISH进行更新的,这不合理,如果任务执行时间比较长,我恰好要看PID,是不是看不到,等到任务结束才能看到

2、修复

Worker端 :
org.apache.dolphinscheduler.plugin.task.api.AbstractCommandExecutor#run
image.png

org.apache.dolphinscheduler.server.worker.runner.TaskCallbackImpl#updateTaskInstanceInfo
image.png

Master端 :
org.apache.dolphinscheduler.server.master.processor.queue.TaskEvent#newUpdatePidEvent
image.png

最后其实由 org.apache.dolphinscheduler.server.master.event.TaskUpdatePidEventHandler 来处理

public void handleTaskEvent(TaskEvent taskEvent) throws TaskEventHandleError {
    int taskInstanceId = taskEvent.getTaskInstanceId();
    int processInstanceId = taskEvent.getProcessInstanceId();

    WorkflowExecuteRunnable workflowExecuteRunnable =
            processInstanceExecCacheManager.getByProcessInstanceId(processInstanceId);
    if (workflowExecuteRunnable == null) {
        sendAckToWorker(taskEvent);
        throw new TaskEventHandleError(
                "Handle task running event error, cannot find related workflow instance from cache, will discard this event");
    }
    Optional<TaskInstance> taskInstanceOptional = workflowExecuteRunnable.getTaskInstance(taskInstanceId);
    if (!taskInstanceOptional.isPresent()) {
        sendAckToWorker(taskEvent);
        throw new TaskEventHandleError(
                "Handle running event error, cannot find the taskInstance from cache, will discord this event");
    }
    TaskInstance taskInstance = taskInstanceOptional.get();
    if (taskInstance.getState().isFinished()) {
        sendAckToWorker(taskEvent);
        throw new TaskEventHandleError(
                "Handle task running event error, this task instance is already finished, this event is delay, will discard this event");
    }

    TaskInstance oldTaskInstance = new TaskInstance();
    TaskInstanceUtils.copyTaskInstance(taskInstance, oldTaskInstance);
    try {
        taskInstance.setStartTime(taskEvent.getStartTime());
        taskInstance.setHost(taskEvent.getWorkerAddress());
        taskInstance.setPid(taskEvent.getProcessId());
        log.info("TaskUpdatePidEventHandler -> {} : pid : {}", taskInstance.getStartTime(), taskEvent.getProcessId());
        // TODO 将PID保存到数据库
        if (!taskInstanceDao.updateById(taskInstance)) {
            throw new TaskEventHandleError("Handle task running event error, update taskInstance to db failed");
        }
        // TODO ack发送回去
        sendAckToWorker(taskEvent);
    } catch (Exception ex) {
        TaskInstanceUtils.copyTaskInstance(oldTaskInstance, taskInstance);
        if (ex instanceof TaskEventHandleError) {
            throw ex;
        }
        throw new TaskEventHandleError("Handle task update pid  event error, update taskInstance to db failed", ex);
    }

}

注意 : 这是主要核心的代码,其他地方的代码自行修改即可(比如说很多测试类)

如感兴趣,点赞加关注,谢谢!!!


journey
32 声望22 粉丝