Hi,
I had the same problem also with kernel 3.4.x (self-compiled, not -beng)
sometimes.
My sympthoms were:
The hole guest was not reachable. No "vserver enter" worked, no ssh or
something else. Also system load increases quickly if the server was
under heavy load. To solve it, I had to reboot the hole machine with all
guests on it. No nice solution....
The problem seems cgroup related. Kernels higher than 3.4.x hangs much
quicker than kernel 3.4. But also 3.4 hangs if the memory limit inside
the guest was reached.
For me the only solution now is to give the guest enough memory so that
it can't reach the limits.
If the hang happens, again I found out that instead of rebooting the
hole host server, it's enough to increase the memory limits on the fly
of the affected guest.
Eg. If you have:
memory.limit_in_bytes (4G) and memory.memsw.limit_in_bytes (8G)
First:
~# echo 10G > /dev/cgroup/$VSNAME/memory.memsw.limit_in_bytes
Second:
~# echo 8G > /dev/cgroup/$VSNAME/memory.limit_in_bytes
Or where ever hou have mounted your cgroups.
The order of the commands must be the same as above.
"memory.memsw.limit_in_bytes" must be increased first.
This solved the hang of the guest immediately and it comes back
to normal operation.
Hope that helps for you, too.
Reagrds
Urban
Am 21.04.2014 20:43, schrieb Tomasz Pajor:
> Thanks for your response. I just downgraded to 3.4-beng, we will see if
> this will fix the issue or not. I'm running 3.13-beng on my other server
> and I don't have this kind of issues.
>
> On 21/04/14 20:23, Sergiusz Pawlowicz wrote:
>> it is pretty normal, downgrade to 3.4-beng and your server will work properly.
>>
>> s.
>
>
Received on Tue Apr 22 08:05:46 2014