[jira] [Commented] (DIRKRB-536) KrbClient leaks sockets

["Kai Zheng (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/DIRKRB-536?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15159936#comment-15159936 ] 

Kai Zheng commented on DIRKRB-536:
----------------------------------

+1 for the patch. Thanks [~Alexander Haskell] for the contribution!

bq. I would also like to propose as a new feature to allow a KrbClient to be able to make multiple requests. As the KrbClient stands now it seems to be a "single shot" tool.
Yes you're right right now it's a single run. This sounds great for a long time running service to reuse a KrbClient instance repeatedly.



> KrbClient leaks sockets
> -----------------------
>
>                 Key: DIRKRB-536
>                 URL: https://issues.apache.org/jira/browse/DIRKRB-536
>             Project: Directory Kerberos
>          Issue Type: Bug
>            Reporter: Alexander Haskell
>            Assignee: Alexander Haskell
>         Attachments: release_socket.patch
>
>
> KrbClient does not close sockets at the end of a request. This is not a problem for KinitTool as the jvm closes the sockets at program termination. A longer running application has to rely on the garbage collector cleaning up open sockets. The most likely location to do this appears to be at the end of DefaultInternalKrbClient.doRequest\{T,S}gt() by calling release() on DefaultInternalKrbClient.transport.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)