GitHub user zhonghaihua opened a pull request:
https://github.com/apache/spark/pull/12258
Ignore task finish for executor lost and removed by driver
## What changes were proposed in this pull request?
Now, when executor is removed by driver with heartbeats timeout, driver
will re-queue the task on this executor and send a kill command to cluster to
kill this executor.
But, in a situation, the running task of this executor is finished and
return result to driver before this executor killed by kill command sent by
driver. At this situation, driver will accept the task finished event and
ignore speculative task and re-queued task.
But, as we know, this executor has removed by driver, the result of this
finished task can not save in driver because the BlockManagerId has also
removed from BlockManagerMaster by driver. So, the result data of this stage is
not complete, and then, it will cause fetch failure. For more details, [link to
jira issues SPARK-14485](https://issues.apache.org/jira/browse/SPARK-14485)
This PR introduce a mechanism to ignore this kind of task finished.
## How was this patch tested?
None.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/zhonghaihua/spark
ignoreTaskFinishForExecutorLostAndRemovedByDriver
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/spark/pull/12258.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #12258
----
commit d491ecd845e41e96aa94f36655860abb0dc70947
Author: zhonghaihua <793507...@qq.com>
Date: 2016-04-08T06:16:05Z
ignore task finished when its executor has already removed by driver
commit c85cf990e84419fa10e3e78e2a32d6e015e46435
Author: zhonghaihua <793507...@qq.com>
Date: 2016-04-08T07:40:07Z
add comment
----
---
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
