Re: OOM didn't save the machine [message #35499 is a reply to message #35492] |
Tue, 31 March 2009 11:36 |
maratrus
Messages: 1495 Registered: August 2007 Location: Moscow
|
Senior Member |
|
|
Hello,
Quote: |
It's possible the "leak"
|
Yes, it might be a leak. Moreover, you said that
Quote: |
Same code runs without problems on non openvz kernels 2.6.22 25 27 for months
|
So, let's try to find out if it's indeed a memory leak.
Quote: |
Now it's happening again, apache process is using 100% cpu, I can't enter the vps beancounters bellow
|
Looks like you have to adjust CPUUNITS/CPULIMITS on the HN. You may read about them in OpenVZ user's guide
http://download.openvz.org/doc/OpenVZ-Users-Guide.pdf
Quote: |
oomguarpages 50345 59833 35236 9223372036854775807 0
|
Oomguarpages exceeded barrier value, so this is likely the reason why a process inside that VE was killed.
http://wiki.openvz.org/Oomguarpages#oomguarpages
Do you have a single VE on the HN? If no, please show the full user_beancounters output.
Next time this issue will occur try to stop the problem VE (hope CPULIMIT/CPUUNITS adjustment will allow to do that) and look at the user_beancounters again. There should be no usage when VE is stopped. Please, also look at the slab state. If the consumption value is decreased. And show please /proc/slabinfo output after and before stopping the problem VE.
|
|
|