[Users] OOM and /etc/vz/oom-groups.conf

[Kir Kolyshkin]

On 10/04/2013 12:05 PM, Joe Doss wrote:
> On 10/03/2013 09:21 PM, Kir Kolyshkin wrote:
>> I went ahead and mocked up something. Please run the attached script on
>> your node
>> and share your results.
> Great script Kir. I gave it a try and I don't think it is calculating
> VSwap correctly:
>
> Swap       available:   12G  allocated:   32Z  oversell: 293109830457468%
>
> At least not on PVC 4.7 node. I for sure do not have 32Z of VSwap set on
> my contaners. Maybe that is due to running it on a PVC hardware node vs
> an OpenVZ hardware node? That seems to be the case.

This is probably because some containers have unlimited (or
near-unlimited) swappages configured, or maybe they are not
VSwap-enabled at all. (I can't figure it out without seeing your
/proc/bc/resources, feel free to email it to me privately so
I can take a look).


> When I ran it on my
> testing OpenVZ hardware node it works fine. One thing I did notice is
> one CT shows the used and peak vswap greater than the limit. Why would
> that be the case? This is on 2.6.32-042stab078.28
>
>         103 1.75G    2G    2G    -    142M  142M  128M    -

This is completely normal -- as http://openvz.org/VSwap tells,
"swap used by a container can exceed swappages.limit, but
is always within sum of physpages.limit and swappages.limit".