[
https://issues.apache.org/jira/browse/NIFI-888?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14708030#comment-14708030
]
Alan Jackoway commented on NIFI-888:
------------------------------------
I made a first pass at this, but it may need more work. I think I got all the
places. However, I wasn't sure about the code in process(). It was closing the
output stream that was passed in. I thought that was wrong except in the case
where the code overrode the stream, so I changed it to conditionally close the
code.
If you are supposed to close an OutputStream passed into process, I need to fix
some of my code...
> EncryptContent for PGP Leaks File Handles
> -----------------------------------------
>
> Key: NIFI-888
> URL: https://issues.apache.org/jira/browse/NIFI-888
> Project: Apache NiFi
> Issue Type: Bug
> Components: Core Framework
> Affects Versions: 0.2.0
> Reporter: Alan Jackoway
> Assignee: Mark Payne
> Priority: Critical
> Fix For: 0.3.0
>
>
> I had some nifi nodes crash last night with errors that suggested file
> handles where exhausted. When I looked into lsof, I found tens of thousands
> of open file handles to a public key I was using in EncryptContent.
> It looks to me like the OpenPGPKeyBasedEncryptor.getPublicKey leaks file
> handles. Because this is called in EncryptContent's validate method, these
> file handles build up even when there is no data flow. You can reproduce by
> the following:
> * Create a PGP public key
> * Set up an EncryptContent processor with the public key for PGP encryption.
> * Do lsof | grep publicKey and see the file handles build up.
> I have a patch. Will post a pull request shortly. I have no idea how to test
> it though other than that I ran it locally and the file handles stopped
> building up when I used the patched code.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
