OpenVZ Forum


Home » Mailing lists » Users » OOM killed process, altough there is enough memory
OOM killed process, altough there is enough memory [message #48274] Wed, 10 October 2012 16:25 Go to next message
Roman Haefeli is currently offline  Roman Haefeli
Messages: 9
Registered: May 2012
Junior Member
Hi all

We're having issues with processes in a container being killed by OOM,
although the hostnode does not have even half of its memory used.
How can it be that there is a failcnt for oomguarpages, although the
barrier wasn't even reached?

This is on Debian Squeeze with the OpenVZ Kernel shipped by Debian.


uname -r
2.6.32-5-openvz-amd64

from dmesg:

[498530.406463] OOM killed process java (pid=4143, ve=2018) exited, free=11882227 gen=17.

from /proc/user_beancounters


uid resource held maxheld barrier limit failcnt
2018: kmemsize 12790390 34388183 34379782 38379782 1255
lockedpages 0 239 1024 1200 6
privvmpages 416902 1559453 2097152 9223372036854775807 0
shmpages 7929 8954 21504 21504 0
dummy 0 0 0 0 0
numproc 184 372 1000 1200 0
physpages 71964 529494 0 9223372036854775807 0
vmguarpages 0 0 2097152 9223372036854775807 0
oomguarpages 71964 529494 2097152 9223372036854775807 1
numtcpsock 43 117 360 360 0
numflock 7 29 188 206 0
numpty 0 1 16 16 0
numsiginfo 0 7 256 256 0
tcpsndbuf 213904 955224 1720320 2703360 0
tcprcvbuf 212992 282952 1720320 2703360 0
othersockbuf 18496 43600 1126080 2097152 0
dgramrcvbuf 0 4360 262144 262144 0
numothersock 214 253 360 360 0
dcachesize 611724 1746754 3409920 3624960 0
numfile 3745 7966 9312 9312 0
dummy 0 0 0 0 0
dummy 0 0 0 0 0
dummy 0 0 0 0 0
numiptent 10 10 128 128 0


Roman
Re: OOM killed process, altough there is enough memory [message #48275 is a reply to message #48274] Wed, 10 October 2012 17:16 Go to previous messageGo to next message
Corin Langosch is currently offline  Corin Langosch
Messages: 18
Registered: March 2012
Junior Member
On 10.10.2012 at 18:25 +0200, Roman Haefeli <reduzent@gmail.com> wrote:
> We're having issues with processes in a container being killed by OOM,
> although the hostnode does not have even half of its memory used.
> How can it be that there is a failcnt for oomguarpages, although the
> barrier wasn't even reached?
>
> This is on Debian Squeeze with the OpenVZ Kernel shipped by Debian.

I can only highly recommend against using the debian openvz kernels. I
always had stability problems with them and a few months ago I
discovered a major bug (
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=655385), causing the
whole system to get instable/ freezed.

For a few months now I'm using the official rhel6 openvz kernels and
convert them into debian packages as described here:
http://wiki.openvz.org/Install_kernel_from_RPM_on_Debian_6.0 Since then
I never had any problems again so far :)

Corin
Re: OOM killed process, altough there is enough memory [message #48276 is a reply to message #48275] Wed, 10 October 2012 20:07 Go to previous message
Roman Haefeli is currently offline  Roman Haefeli
Messages: 9
Registered: May 2012
Junior Member
On Mit, 2012-10-10 at 19:16 +0200, Corin Langosch wrote:
> On 10.10.2012 at 18:25 +0200, Roman Haefeli
> <reduzent@gmail.com> wrote:
> > We're having issues with processes in a container being killed by OOM,
> > although the hostnode does not have even half of its memory used.
> > How can it be that there is a failcnt for oomguarpages, although the
> > barrier wasn't even reached?
> >
> > This is on Debian Squeeze with the OpenVZ Kernel shipped by Debian.
>
> I can only highly recommend against using the debian openvz kernels. I
> always had stability problems with them and a few months ago I
> discovered a major bug
> ( http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=655385), causing
> the whole system to get instable/ freezed.
>
> For a few months now I'm using the official rhel6 openvz kernels and
> convert them into debian packages as described here:
> http://wiki.openvz.org/Install_kernel_from_RPM_on_Debian_6.0 Since
> then I never had any problems again so far :)

Thanks for your comments. It's not the first time I experienced oddities
with the Debian OpenVZ kernel. I'll definitely check out the RHEL6
kernel.

Also, knowing that what I reported is _not_ expected behavior is
valuable for me to know. I started to think that I might have
misconfigured the CTs due to a lack of understanding the mem
limits/guarantees.

Roman
Previous Topic: reboot in CT still does not work
Next Topic: Memory size confusion
Goto Forum:
  


Current Time: Tue Mar 19 05:28:38 GMT 2024

Total time taken to generate the page: 0.02411 seconds