So it turns out that if your /dev/shm fills up (or wherever uml is
keeping the memory files), it crashes the machine that is requesting
more memory silently. I increased the size of my /dev/shm by about 4x,
and the problem has been solved. Still, UML should write something to
syslog on the host or guest...
Jeff Dike wrote:
> On Fri, Dec 08, 2006 at 02:21:37PM -0500, Jonas Meyer wrote:
>> Hello folks. I'm hoping someone online could help with these random
>> crashes we've been having. Guests keep randomly crashing for no good
>> reason, and they leave no logs. I've checked the logs on host and on
>> the guest, and I get nothing. They just randomly terminate in a
>> non-repeatable manner. Obviously this is a little frustrating,
>> especially since this is a production environment.
> I've heard from one other person about this sort of thing.
> If you have /proc/sys/kernel/print-fatal-signals on the host, can you turn
> it on? We need some information to go on, and right now, there's nothing.