We're aware of problems in our implementation of connection disposal on time out and we're working to implement fixes for common error scenarios before we ship 1.4.0 and a hot fix for 1.3.x - while not a complete solution for the unbounded connection pool that you're seeing, this should alleviate some of the pain, especially on idle socket timeouts.
I have a few questions for you so we can better direct our fixes and functionality (both right now and going forward): What are the data access patterns you're primarily using? Are you using the streaming APIs? Are you getting rid of/fully enumerating collections before moving on? How are you handling errors processing individual items in the IEnumerable? (This is primarily a concern if you're using the streaming interfaces) --- Jeremiah Peschka - Founder, Brent Ozar Unlimited MCITP: SQL Server 2008, MVP Cloudera Certified Developer for Apache Hadoop On Thu, Jul 11, 2013 at 7:46 AM, Amir Halatzi <[email protected]>wrote: > We’ve tried that. The result was 3K open connections per Riak node, and > as you can guess – this is not an improvement.**** > > ** ** > > Do you have other recommendations?**** > > ** ** > > Thanks,**** > > Amir**** > > ** ** > > ** ** > > *From:* OJ Reeves [mailto:[email protected]] > *Sent:* 11 July 2013 17:31 > *To:* Amir Halatzi > *Cc:* Jeremiah Peschka; [email protected] > > *Subject:* Re: CorrugatedIron & Idle Connection**** > > ** ** > > This is a discussion we really need to have. Thanks for opening the Github > issue.**** > > ** ** > > In the short term, you can "turn off" pooling by setting the pool size to > zero. This will mean all the connections are done "on the fly".**** > > ** ** > > On Fri, Jul 12, 2013 at 12:26 AM, Amir Halatzi <[email protected]> > wrote:**** > > Just did: https://github.com/DistributedNonsense/CorrugatedIron/issues/153 > **** > > **** > > Still, if we’d like to manage it ourselves - how would you recommend going > about it? Should we interfere with the way you handle the connection, or > just retry the action on ‘network failure’?**** > > **** > > BR,**** > > Amir**** > > **** > > *From:* Jeremiah Peschka [mailto:[email protected]] > *Sent:* 11 July 2013 17:11 > *To:* Amir Halatzi > *Cc:* [email protected] > *Subject:* Re: CorrugatedIron & Idle Connection**** > > **** > > We're in the process of figuring out how to best handle killing off idle > connections inside of CorrugatedIron. If you can head over to > https://github.com/DistributedNonsense/CorrugatedIron/issues and file a > new bug report, that would be fantastic.**** > > **** > > We've avoided killing off connections until now because we don't want to > introduce locking into our connection management. Truthfully, we haven't > run into a bug report where the network team was killing off connections.* > *** > > **** > > So, anyway, if you can submit a bug report we'll get on it. It won't make > it out for the 1.4.0 release, but we can put that in a point release.**** > > > **** > > ---**** > > Jeremiah Peschka - Founder, Brent Ozar Unlimited**** > > MCITP: SQL Server 2008, MVP**** > > Cloudera Certified Developer for Apache Hadoop**** > > **** > > On Thu, Jul 11, 2013 at 6:59 AM, Amir Halatzi <[email protected]> > wrote:**** > > Hi all, > > Trying to optimize our application connection pool, we encountered and > issue with Corrugated Iron and identifying idle/reset connections. Namely, > if our network infrastructure (firewall/proxy/etc) closes an idle > connection - we're getting a 'cluster is offline' error code. > > I saw that this is a feature in the making (for the past 8 months), but I > was wondering if any of you found a way around it? > > Thank you, > Amir Halatzi, > ooVoo > > > _______________________________________________ > riak-users mailing list > [email protected] > http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com**** > > **** > > > _______________________________________________ > riak-users mailing list > [email protected] > http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com**** > > > > **** > > ** ** > > -- > > OJ Reeves**** > > +61 431 952 586 > http://buffered.io/**** >
_______________________________________________ riak-users mailing list [email protected] http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
