On Thu, Jun 26, 2014 at 10:03 PM, Anand Kumar, Karthik < karthik.anandku...@classmates.com> wrote:
> We run postgres 9.3.3 on Centos 6.3, kernel 2.6.32-431.3.1. Every once in > a while, we see postgres processes spinning on semop: > it might be long shot, but when we had problems with lots of backends "sitting" in semop, it was solved by: 1. disabling zone_reclaim (echo 0 > /proc/sys/vm/zone_reclaim_mode) 2. disabling transparent hugepage support - this has various names on different kernel/distributions, but "find /sys | grep -i transparent.*hugepage.*enable" will find it, and then just echo never there. depesz
