On Wed, Apr 30, 2014 at 05:18:30PM +0000, Fiedler Roman wrote:
> Hello,
> I'm trying to solve following problem, perhaps someone has an
> useful idea on that:
> There is a vserver host running about 50 guests. In one guest,
> a python process crashes without writing any logs by itself -
> it might just be a temporary job crashing and the caller does
> not log the event either. But the kernel logs the fatal signal
> including process name and pid but not the context, so I know,
> that something is crashing but not in which guest.
You could add the context number to the kernel message.
> Since all guests run numerous python processes, is there an
> easy way to find out, in which context the crash might have
> occurred so that I have to inspect only one of them?
What kernel/patch are you running?
> Another way could be to enable core dumps for all current and
> future processes and see, where a core dump is written. Is this
> somehow possible?
You could probably do that by overriding the core dump
limits in a custom kernel.
Best,
Herbert
> Any comments welcome,
> Roman
Received on Thu May 1 02:24:49 2014