KnightChess opened a new issue #3900:
URL: https://github.com/apache/incubator-dolphinscheduler/issues/3900
Forgive my poor english
**Describe the question**
A clear and concise description of what the question is.
When kill yarn task, dolphinscheduler may shows the kill is successful, but
the tasks may still running on the yarn cluster
* * *
当 kill yarn 类型的任务的时候，dolphinscheduler 上显示任务状态 kill 成功，但实际任务可能还在 yarn
集群上运行，没有 kill 掉
**Which version of DolphinScheduler:**
-[1.3.3-release]
**Additional context**
Add any other context about the problem here.
code in
org.apache.dolphinscheduler.server.utils.ProcessUtils.cancelApplication
String appid = appIds.get(appIds.size() - 1);
get the last appId，If the task has multiple stages，and hive.exec.parallel =
true，stages will be executed in parallel if there is no dependency between
them. If the last job for stage is successful(last appId), but other is
running, the task will kill fail in yarn cluster, but dolphinscheduler show
success.
* * *
上面的代码是获取 appId 列表中的最后一个进行
kill，如果这个任务有多个stage，并且stage之间没有依赖啊，hive.exec.parallel = true 设置可以并行执行，就会存在最后一个
appId 对应的 job 执行成功，但是中间或者还存在没有执行完的，如果这时候在 dolphinscheduler 将任务 kill 掉，那么 kill
最后一个 appId 没法在集群上 kill 掉任务的，因为最后一个 appId 对应的 job 已经成功，但是还存在其它并行执行的任务
**Requirement or improvement**
- Please describe about your requirements or improvement suggestions.
I haven't tried it on dolphinscheduler，when study dolphinscheduler code,
this killYarnJob code is similar to our self-developed scheduler，we meet this
question in our system
* * *
我没有在 dolphinscheduler 上试过，但是我在看相关代码的时候，发现和我们自研的调度系统 killYarnJob
的方式是差不多一样得，都是获取最后一个 appId 进行 kill，就会产生这个问题
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
