Re: Postmaster only takes 4-5% CPU

Under some circumstances, on my (somewhat) older 7.2 box running the 2.4.9
kernel, the kswapd and bdflush daemons will start working overtime when
there are accesses by windows users via samba to look through large
directories.

This makes the box QUITE slow and pokey, and all the tasks that are
running besides kswapd / bdflush are in D state mostly, while the swap
daemon eats 0 to 100% CPU depending on some change in state in samba I've
not bothered tracking down.

I understand a newer kernel might be a good thing for me, so eventually
I'll do that.  For now I just put the Windows Samba people on another box.
They both go through the the same NAS, and it keeps the load off the main
server.

So, look for other processes that might be doing something like scanning a
directory of 40,000 files or some such.

On Thu, 29 May 2003, Ruben wrote:

> Thanks for your answers Jan, Arjen, Shridar and Scott:
>
> Postmaster is back taking lots of CPU when needed. I guess you were
> right about Postgres waiting for some I/O activity. Anyway, I was not
> able to see what process was responsible for this, since top showed
> nothing remarkable.
>
> This is a new top while running a VACUUM of the database:
>
>    8:51pm  up  6:30,  3 users,  load average: 0,45, 0,11, 0,04
> 93 processes: 89 sleeping, 4 running, 0 zombie, 0 stopped
> CPU states:  0,1% user,  1,1% system,  0,0% nice, 98,6% idle
> Mem:  1031012K av, 1019612K used,   11400K free,  0K shrd,   44896K buff
> Swap: 2040244K av,    1716K used, 2038528K free          894008K cached
>
>    PID USER     PRI  NI  SIZE  RSS SHARE STAT %CPU %MEM   TIME COMMAND
>   1673 postgres  15   0  5120 5120  3196 R     0,9  0,4   0:02 postmaster
>   1683 root      15   0  1044 1044   840 R     0,1  0,1   0:00 top
>      1 root      15   0   476  476   424 S     0,0  0,0   0:04 init
>      2 root      15   0     0    0     0 SW    0,0  0,0   0:00 keventd
>      3 root      15   0     0    0     0 SW    0,0  0,0   0:00 kapmd
>      4 root      34  19     0    0     0 SWN   0,0  0,0   0:00
> ksoftirqd_CPU0
>      5 root      15   0     0    0     0 SW    0,0  0,0   0:00 kswapd
>      6 root      15   0     0    0     0 SW    0,0  0,0   0:00 bdflush
>      7 root      15   0     0    0     0 SW    0,0  0,0   0:00 kupdated
>      8 root      25   0     0    0     0 SW    0,0  0,0   0:00 mdrecoveryd
>     12 root      15   0     0    0     0 SW    0,0  0,0   0:00 kjournald
>     66 root      15   0     0    0     0 SW    0,0  0,0   0:00 khubd
>    266 root      15   0     0    0     0 SW    0,0  0,0   0:00 kjournald
>    267 root      15   0     0    0     0 SW    0,0  0,0   0:00 kjournald
>    268 root      15   0     0    0     0 SW    0,0  0,0   0:05 kjournald
>    269 root      15   0     0    0     0 SW    0,0  0,0   0:00 kjournald
>    547 root      15   0     0    0     0 SW    0,0  0,0   0:00 eth0
>
> Cheers, Ruben.
>
>
> Jan Poslusny wrote:
> > Did not your postmaster wait for harddisk ?
> >
> > Ruben wrote:
> >
> >> Hi all:
> >>
> >> When I run a (big) query from pgsql, postmaster used to take almost
> >> 100% of CPU. Now, with aparent no reason, Postmaster only takes 4-5%
> >> CPU and the same query takes really longer to finish.
> >>
> >> I have tried renicing -20 postmaster but it still only takes a small
> >> percentage of CPU, while 90% is idle.
> >>
> >> This is a "top" screen while running the query:
> >>
> >>   2:27pm  up 5 min,  2 users,  load average: 0,80, 0,26, 0,08
> >> 74 processes: 72 sleeping, 2 running, 0 zombie, 0 stopped
> >> CPU states:  3,1% user,  2,1% system,  0,0% nice, 94,6% idle
> >> Mem:  1031012K av,  246040K used,  784972K free,       0K shrd,
> >> 11496K buff
> >> Swap: 2040244K av,       0K used, 2040244K free
> >> 196916K cached
> >>
> >>   PID USER     PRI  NI  SIZE  RSS SHARE STAT %CPU %MEM   TIME COMMAND
> >>  1165 postgres   0 -20  3476 3476  2912 D <   4,5  0,3   0:02 postmaster
> >>   268 root      15   0     0    0     0 SW    0,1  0,0   0:00 kjournald
> >>
> >> Any idea of how can I get postmaster to eat as much CPU as possible?
> >>
> >> Thanks a lot! Ruben.
> >>
> >>
> >> ---------------------------(end of broadcast)---------------------------
> >> TIP 4: Don't 'kill -9' the postmaster
>
>
> ---------------------------(end of broadcast)---------------------------
> TIP 2: you can get off all lists at once with the unregister command
>     (send "unregister YourEmailAddressHere" to majordomo@postgresql.org)
>