On 12/15/2011 2:28 PM, Kyle Waters wrote: > I am running Bucardo in a simple master slave system. I have bucardo > running on the slave server. Periodically(couple times a day) it seems > to hang. If I connect into the master system and run top I can see a > postmaster process that has been running for hours and using 99% of a > CPU. This seems to affecting database performance. Any suggestions on > how to diagnose and resolve this issue? > > Kyle > > /* > PLUG: http://plug.org, #utah on irc.freenode.net > Unsubscribe: http://plug.org/mailman/options/plug > Don't fear the penguin. > */ > > > ----- > No virus found in this message. > Checked by AVG - www.avg.com > Version: 2012.0.1890 / Virus Database: 2108/4682 - Release Date: 12/15/11 > > I've seen one of two approaches taken to this problem. The obvious and tedious process of researching the thread and dealing with the problem is probably your best choice. Option number two which i suggest no matter how good you think you are at server management is to build/install a watchdog. I usually recommend building one if you have the skill because most of the watchdog software out there comes with more overhead and cost that is strictly necessary. Basically you have a cron look at your process list every 5-10 minutes, diagnostic versions might have to run more often. ps -ef is a decent standard command line regex parse-able command. You can either have it kill or just document rouge threads. I usually have my watchdogs log and kill them if it's a production server just because clients don't care why the server is slow they just get mad when it is. On my development servers i tend to let em run and just log activity for the programmers to review later.
Jason /* PLUG: http://plug.org, #utah on irc.freenode.net Unsubscribe: http://plug.org/mailman/options/plug Don't fear the penguin. */
