hc-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Oleg Kalnichevski (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HTTPCLIENT-814) MultiThreadedHttpConnectionManager deadlocks when reusing socket connections
Date Fri, 09 Jan 2009 15:26:59 GMT

    [ https://issues.apache.org/jira/browse/HTTPCLIENT-814?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12662398#action_12662398

Oleg Kalnichevski commented on HTTPCLIENT-814:

I suspect rather strongly your application is leaking connections by not returning them back
to the pool. Connections are not supposed just to pile up, unless your application is always
connecting to a new host or not releasing existing connections back to the pool.

BTW, consider upgrading to 4.0 which has a massively better connection management architecture.


> MultiThreadedHttpConnectionManager deadlocks when reusing socket connections
> ----------------------------------------------------------------------------
>                 Key: HTTPCLIENT-814
>                 URL: https://issues.apache.org/jira/browse/HTTPCLIENT-814
>             Project: HttpComponents HttpClient
>          Issue Type: Bug
>          Components: HttpClient
>    Affects Versions: 3.1 Final
>         Environment: Tomcat 6.18 running on a Windows 2003 server
>            Reporter: Will Franco
>             Fix For: 3.1.1
> During our Load Testing we encountered that the MultiThreadedHttpConnectionManager deadlocks
when reusing socket connections.   These socket connections remain in the pool in a CLOSE_WAIT
state, and the pool is unable to reuse them or create new ones because the maximum number
of connection has already been reached, thus deadlocking.
> Let me explain briefly how this library is being used and the temporary work around that
we are using.  However, the correct solution would be to use the connection pool.
> Usage:
> 1) The HttPClient is being used to handle connections to another server from our Web
Application.  So far, for every request our Web Apps processes we have to make at least one
request to another server using the HttpClient.
> 2) Our load tester, loads our web app with thousands of request within the first 30 seconds,
and maintains such high volume of requests for at least 24 hours
> Work Around:
> The current work around we implemented is to have different instances of the HttpClient
per request, and we shutdown the Connection Manager once we complete the transaction with
the other server.
> The next work around we will be implementing is to called the closeIdleConnections, this
solution will remove the connections that can pontentialy deadlock the connection manager
allowing it to allocate new ones.
> Please email me to willfranco@gmail.com if you need any further information regarding
this issue.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

To unsubscribe, e-mail: dev-unsubscribe@hc.apache.org
For additional commands, e-mail: dev-help@hc.apache.org

View raw message