-
Notifications
You must be signed in to change notification settings - Fork 137
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Bug] problems of RssMapOutputCollector #715
Comments
Is it an occasional case? |
An inevitable case |
Could you raise a pr to fix this issue? |
jerqi
pushed a commit
that referenced
this issue
May 19, 2023
…not closed (#882) ### What changes were proposed in this pull request? The container does not exit because shuffleclient is not closed ### Why are the changes needed? For #715 1.The process does not exit after the maptask or reducetask execution is complete. The reason is that ShuffleWriteClient has a thread pool that does not close when the task completes. So turning off ShuffleWriteClient can solve this problem. 2.How do I recreate this scene? Initialize a small cluster and submit an mr Task whose requested resources exceed the total resources in the cluster. We can see that all tasks have completed execution without quitting until the timeout time exceeds 60 seconds(mapreduce.task.exit.timeout). The appmaster requests the nodemanager to kill the corresponding container. The nodemanager logs are as follows `2023-03-12 13:56:45,901 INFO [AsyncDispatcher event handler] org.apache.hadoop.mapreduce.v2.app.job.impl.TaskAttemptImpl: Diagnostics report from attempt_1676901654399_1653119_m_000070_0: [2023-03-12 13:56:44.909]Container killed by the ApplicationMaster. [2023-03-12 13:56:44.921]Sent signal OUTPUT_THREAD_DUMP (SIGQUIT) to pid 45556 as user tc_infra for container container_e304_1676901654399_1653119_01_000072, result=success [2023-03-12 13:56:44.985]Container killed on request. Exit code is 143 [2023-03-12 13:56:45.403]Container exited with a non-zero exit code 143. ` ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? existing UTs. Co-authored-by: zhaobing <zhaobing@zhihu.com>
closed by #882 |
jerqi
pushed a commit
that referenced
this issue
May 19, 2023
…not closed (#882) ### What changes were proposed in this pull request? The container does not exit because shuffleclient is not closed ### Why are the changes needed? For #715 1.The process does not exit after the maptask or reducetask execution is complete. The reason is that ShuffleWriteClient has a thread pool that does not close when the task completes. So turning off ShuffleWriteClient can solve this problem. 2.How do I recreate this scene? Initialize a small cluster and submit an mr Task whose requested resources exceed the total resources in the cluster. We can see that all tasks have completed execution without quitting until the timeout time exceeds 60 seconds(mapreduce.task.exit.timeout). The appmaster requests the nodemanager to kill the corresponding container. The nodemanager logs are as follows `2023-03-12 13:56:45,901 INFO [AsyncDispatcher event handler] org.apache.hadoop.mapreduce.v2.app.job.impl.TaskAttemptImpl: Diagnostics report from attempt_1676901654399_1653119_m_000070_0: [2023-03-12 13:56:44.909]Container killed by the ApplicationMaster. [2023-03-12 13:56:44.921]Sent signal OUTPUT_THREAD_DUMP (SIGQUIT) to pid 45556 as user tc_infra for container container_e304_1676901654399_1653119_01_000072, result=success [2023-03-12 13:56:44.985]Container killed on request. Exit code is 143 [2023-03-12 13:56:45.403]Container exited with a non-zero exit code 143. ` ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? existing UTs. Co-authored-by: zhaobing <zhaobing@zhihu.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Code of Conduct
Search before asking
Describe the bug
The connection between the client and shufflerserver is not closed. As a result, the maptask container does not exit.
1.map container does not exit when reduce is running
2.When cluster resources are used up and some maps are not allocated, am waits for a one-minute timeout, kills the completed map container, and allocates resources to other maps
error log like:
2023-03-11 02:42:16,826 INFO [Ping Checker] org.apache.hadoop.yarn.util.AbstractLivelinessMonitor: Expired:attempt_1676901654399_1531374_m_000190_0 Timed out after 60 secs
In the close method of the RssMapOutputCollector, closing the shuffle client solves this problem
Affects Version(s)
master
Uniffle Server Log Output
No response
Uniffle Engine Log Output
No response
Uniffle Server Configurations
No response
Uniffle Engine Configurations
No response
Additional context
No response
Are you willing to submit PR?
The text was updated successfully, but these errors were encountered: