github-actions[bot] commented on issue #18247: URL: https://github.com/apache/dolphinscheduler/issues/18247#issuecomment-4428291225
### Search before asking - [x] I had searched in the [issues](https://github.com/apache/dolphinscheduler/issues?q=is%3Aissue) and found no similar issues. ### What happened When performing a forced success operation on a worker node instance through the public void forceTaskSuccess(User loginUser, long projectCode, Integer taskInstanceId) method, if the node is the only failed/kill node in the workflow, the workflow instance should be automatically marked as SUCCESS. But in reality the workflow instance is never promoted to SUCCESS. **Example:** I now have two worker node instances, one successful and one failed <img width="907" height="428" alt="Image" src="https://github.com/user-attachments/assets/850952f3-553d-4962-9f25-0681ecb0fa68"; /> Now I force this failed worker node instance to succeed <img width="1587" height="284" alt="Image" src="https://github.com/user-attachments/assets/cc1884d7-248c-4ed5-b2c9-1501a322a928"; /> According to the logic of processService.forceWorkflowInstanceSuccessByTaskInstanceId(task);, the corresponding workflow instance should be set to success. <img width="1266" height="280" alt="Image" src="https://github.com/user-attachments/assets/d10f552f-62c6-40d2-a668-2a241db52ece"; /> But the workflow instance status is not updated. ### What you expected to happen <img width="1266" height="280" alt="Image" src="https://github.com/user-attachments/assets/8fb69d49-a3f5-443c-a7ed-cf1f2be2e428"; /> The status of this workflow instance was successfully updated. ### How to reproduce In the forceTaskSuccess method, the task instance status is first updated to FORCED_SUCCESS and persisted to the database, and then processService.forceWorkflowInstanceSuccessByTaskInstanceId(task); is called. <img width="1117" height="540" alt="Image" src="https://github.com/user-attachments/assets/9174052b-8ee3-494f-8053-6710a210792d"; /> Within processService.forceWorkflowInstanceSuccessByTaskInstanceId, the task instance list will be re-queried from the database: List<TaskInstance> validTaskList = taskInstanceDao.queryValidTaskListByWorkflowInstanceId(workflowInstance.getId()); At this time, the status of the task instance in the database is FORCED_SUCCESS (its own write operations in the same transaction should be read immediately), and it no longer meets the filtering conditions of isFailure() || isKill(): List<Integer> failTaskList = validTaskList.stream() .filter(instance -> instance.getState().isFailure() || instance.getState().isKill()) .map(TaskInstance::getId).collect(Collectors.toList()); if (failTaskList.size() == 1 && failTaskList.contains(task.getId())) { // Never able to enter - failTaskList does not contain tasks that have been forced to succeed } Since the task that is forced to succeed is no longer in the failure/kill state, the situation of failTaskList is: - If it is the only failed node → failTaskList is empty → condition is not met - If there are other failed nodes → failTaskList does not contain the current task → failTaskList.contains(task.getId()) returns false Therefore, a workflow instance can never be marked SUCCESS. **Fix:** Change the calling order: call forceWorkflowInstanceSuccessByTaskInstanceId() before updating the task instance status, so that the original failure/kill status can be read during internal query: processService.forceWorkflowInstanceSuccessByTaskInstanceId(task); // change the state of the task instance task.setState(TaskExecutionStatus.FORCED_SUCCESS); task.setEndTime(new Date()); int changedNum = taskInstanceMapper.updateById(task); if (changedNum <= 0) { throw new ServiceException(Status.FORCE_TASK_SUCCESS_ERROR); } ### Anything else _No response_ ### Version dev ### Are you willing to submit PR? - [ ] Yes I am willing to submit a PR! ### Code of Conduct - [x] I agree to follow this project's [Code of Conduct](https://www.apache.org/foundation/policies/conduct) -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
