Hi Christoph,

thanks for your response

Christoph Lameter wrote:
> On Thu, 10 Feb 2011, Peter Kruse wrote:
> 
>> today one of our servers went berserk and produced literally 353
>> page allocation failures in 7 minutes until it was reset
>> (sysrq was still working).  I attach one of them as an example.
>> The failures happened for different processes ranging from
>> sshd, top, java, tclsh, ypserv, smbd, portmap, kswapd to Xvnc4.
>> I already reported about an incidence with this server here:
>> https://lkml.org/lkml/2011/1/19/145
> 
> Atomic allocations are failing there? gfpmask = 0x20?
> 
>> we have set vm.min_free_kbytes = 2097152 but the problem
>> obviously did not go away.
> 
> 2GB of reserves? How much memory does your system have?

48GB

> 
>> Please anybody, what is the cause of these failures?
> 
> Could you post the entire messages from the kernel log? We need the OOM
> info to figure out more about the problem.
> 

I attach one of the call traces, or would it be better if I send the
kern.log (about 6MB)?

Best regards,

   Peter