[
https://issues.apache.org/jira/browse/KNOX-1430?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16598027#comment-16598027
]
Guang Yang commented on KNOX-1430:
----------------------------------
This might not be the root cause, because the file descriptor issue went away
when I upgrade jetty to 9.3.10 on Knox 1.1.0. But there are some orphan http
connections in our system, as we see the http connection count increases
gradually and eventually use up all the connections in the pool. But not sure
whether this is related.
> Potential output stream handle leak when copyLarge files in streamResponse
> --------------------------------------------------------------------------
>
> Key: KNOX-1430
> URL: https://issues.apache.org/jira/browse/KNOX-1430
> Project: Apache Knox
> Issue Type: Bug
> Components: Server
> Affects Versions: 0.14.0, 1.1.0
> Reporter: Guang Yang
> Assignee: Guang Yang
> Priority: Major
> Fix For: 1.2.0
>
> Attachments: KNOX-1430.patch
>
>
> When Knox streams a large response, like download a 2G file through webhdfs.
> The output stream handle will leak if there is an exception when copy the
> large file. This fix move outputStream.close() to the finally block.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
