[
https://issues.apache.org/jira/browse/HADOOP-18313?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17559032#comment-17559032
]
wujinhu commented on HADOOP-18313:
----------------------------------
Thanks [[email protected]] It's a good suggestion for our OSS users. Flink
users depend on hadoop-aliyun module to write to OSS. Their workers run OOM
because workers do not exit. This fix can resolve the issue.
> AliyunOSS: AliyunOSSBlockOutputStream should not mark the temporary file for
> deletion
> -------------------------------------------------------------------------------------
>
> Key: HADOOP-18313
> URL: https://issues.apache.org/jira/browse/HADOOP-18313
> Project: Hadoop Common
> Issue Type: Bug
> Components: fs/oss
> Affects Versions: 2.10.2, 3.2.3, 3.3.3
> Reporter: wujinhu
> Assignee: wujinhu
> Priority: Major
> Labels: pull-request-available
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> AliyunOSSBlockOutputStream buffers data in local directory before uploading
> to OSS. It uses LocalDirAllocator.createTmpFileForWrite which will create a
> temp file on this directory and invoke DeleteOnExitHook.add to add the path
> to a set and will be deleted on VM exit through a shutdown hook. However, the
> size of set will be growing if the VM does not exit.
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
