> this is the context when every thing is fine, because we until this thread
> we didn't realize that the server swaps. Because shortly before the crash
> our monitoring reported 0 bytes swap used. And at exploding time there
> were 4019584k of swap in use, so why can this happens when ech vserver
> has a maximum of 3 GB rss settings? In the dmesh output there was only on
> process in oom killer. Is there maybe a bug in oomkiller allocating all
> memroy and swap on the host?
Do you have kernel logs from when oomkiller was triggered?
I would be very interested in seeing the events and actual activity of
the oomkiller because a co-worker and I have been chasing something
very similar.
-- KyleReceived on Thu Jan 21 18:44:09 2010