On Fri, Oct 23, 2015 at 12:45 PM, Jamie Koceniak
<jkoceniak@mediamath.com> wrote:
> Hi,
>
> We just had the performance problem again today.
> Here is some of the top output. Unfortunately, we don't have perf top installed.
>
> top - 16:22:16 up 29 days, 13:00, 2 users, load average: 164.63, 158.62, 148.52
> Tasks: 1369 total, 181 running, 1188 sleeping, 0 stopped, 0 zombie
> %Cpu(s): 6.2 us, 0.7 sy, 0.0 ni, 93.1 id, 0.0 wa, 0.0 hi, 0.0 si, 0.0 st
> MiB Mem: 2068265 total, 433141 used, 1635124 free, 586 buffers
> MiB Swap: 7812 total, 0 used, 7812 free, 412641 cached
>
> PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
> 81745 postgres 20 0 65.7g 51m 34m R 101 0.0 0:09.20 postgres: user1 db 0.0.0.2(52307) SELECT
> 81782 postgres 20 0 65.7g 51m 34m R 101 0.0 0:08.50 postgres: user1 db 0.0.0.3(44630) SELECT
> 81797 postgres 20 0 65.7g 51m 34m R 101 0.0 0:08.03 postgres: user1 db 0.0.0.6(60752) SELECT
<snip>
ok, this rules out iowait.
load is 160+. system is reporting 6.2%user, 93.1%idle, 0 iowait.
This is very odd.
*) how many processors do you have?
*) Can we have more details about the hardware platform?
*) Is this system virtualized? If so, what solution?
we need a perf top and a capture of 'vmstat 1' for context switches
merlin