Github user lw-lin commented on the issue:

    https://github.com/apache/spark/pull/13685
  
    Hi @squito thanks for the comments!
    
    > how you'd have a TaskKilledException, but without setting the task to 
`killed`
    
    This can be reproduced when, a task gets 
killed([Executor#L235~L252](https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/executor/Executor.scala#L235~L252))
 before it can be 
deserialized([Executor#L253](https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/executor/Executor.scala#L253)).
 Then in 
[TaskRunner.kill()](https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/executor/Executor.scala#L211),
 `task` is still `null`, so only `taskRunner.killed` is set to `true` but not 
`task.killed`.
    
    The updated test reproduces this. In the test, we use `mockExecutorBackend` 
to trigger `executor.killAllTasks(true)` at Executor#L240; the program should 
run in this order:
    ```
    
+-----------------------------+----------------------------------------------+
    |      main test thread       |      worker thread                          
 |
    
+-----------------------------+----------------------------------------------+
    |    executor.launchTask()    |                                             
 |
    |                             | TaskRunner.run() begins                     
 |
    |                             |          ...                                
 |
    |                             | execBackend.statusUpdate  // 1st time, 
#L240 |
    | executor.killAllTasks(true) |                                             
 |
    |                             |          ...                                
 |
    |                             |  task = ser.deserialize   // #L253          
 |
    |                             |          ...                                
 |
    |                             | execBackend.statusUpdate  // 2nd time, 
#L365 |
    |                             |          ...                                
 |
    |                             |   TaskRunner.run() ends                     
 |
    |       check results         |                                             
 |
    
+-----------------------------+----------------------------------------------+
    ```
    
    Then:
    
    <table>
    <tr>
        <td align="center"><strong></strong></td>
        <td align="center"><strong>prior to this patch</strong></td>
        <td align="center"><strong>after this patch</strong></td>
    </tr>
    <tr>
        <td align="center">testFailedReason</td>
        <td align="center">ExceptionFailure</td>
        <td align="center">TaskKilled</td>
    </tr>
    <tr>
        <td align="center">taskState</td>
        <td align="center">FAILED</td>
        <td align="center">KILLED</td>
    </tr>
    </table>


---
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.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org

Reply via email to