[
https://issues.apache.org/jira/browse/HBASE-7515?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13548735#comment-13548735
]
Elliott Clark commented on HBASE-7515:
--------------------------------------
v4 looks better to me than v5. I don't really like adding synchronized bits
just for cosmetic code reasons.
However JD had a good idea. Closing the reader that throws an exception could
be pushed up into to storefile.createReader(). Then what happens is that any
store file that errors out has already cleaned up after itself. Then we just
have to close the other store files that didn't error out. That makes things a
little cleaner in my opinion.
> Store.loadStoreFiles should close opened files if there's an exception
> ----------------------------------------------------------------------
>
> Key: HBASE-7515
> URL: https://issues.apache.org/jira/browse/HBASE-7515
> Project: HBase
> Issue Type: Bug
> Affects Versions: 0.94.3
> Reporter: Jean-Daniel Cryans
> Assignee: Ted Yu
> Fix For: 0.96.0, 0.94.5
>
> Attachments: 7515.txt, 7515-v2.txt, 7515-v3.txt, 7515-v4.txt,
> 7515-v5.txt
>
>
> Related to HBASE-7513. If a RS is able to open a few store files in
> {{Store.loadStoreFiles}} but one of them fails like in 7513, the opened files
> won't be closed and file descriptors will remain in a CLOSED_WAIT state.
> The situation we encountered is that over the weekend one region was bounced
> between >100 region servers and eventually they all started dying on "Too
> many open files".
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
