Nothing appreared in /var/log/messages about the oom killer on both the boxes. So could be more like an firewall issue, will look into this and let you know. Current firewall setting that we have are:
On Tue, Aug 26, 2014 at 9:57 AM, Scott Marlowe <scott.marlowe@gmail.com> wrote: > On Tue, Aug 26, 2014 at 9:27 AM, Dhruv Shukla <dhruvshukla82@gmail.com> wrote: > >> On viewing server logs on postgres server found the error as >> FATAL: connection to client lost >> >> There has been some improvements, like few of the getting stuck DB functions >> ran good. But the main ones are still failing. > > This sounds like a firewall issue to me, but it could be a backend > running out of memory and the oom killer in linux killing it. I > believe oom killer logs to /var/log/syslog or /var/log/messages. Got > anything in those logs about that time for the oom killer? > > If not, it could be a firewall, if you have one, between the app and > db boxes. If that's the case you can stop it from happening by > lowering the tcp keepalive settings.
Please note that the oom killer is deadly to a postgres box, as it can kill the postmaster which is a bad thing. It should be disabled by setting vm.overcommit_memory=2 in /etc/sysctl.conf and running sysctl -p -- To understand recursion, one must first understand recursion.