XFS hangs with XFS: possible memory allocation deadlock in kmem_alloc

[Michael Meier]

On 03/07/2015 03:07 PM, Brian Foster wrote:
> Thanks for the data. Some notes from the backtraces in the first
> instance:

Thank you for the quick reply.
I'm not sure if the first instance is the most representative: It was
very short - only one message was logged and then everything was fine
again. The later one starting at 00:48 in the logs however was long
enough to make our nagios complain.

> Considering this is a large memory box (64g), I wonder if some vm tuning
> might help mitigate this behavior..? For example, increase
> /proc/sys/vm/min_free_kbytes in hopes of allowing more memory for these
> allocations when under pressure, or tune down the
> dirty_ratio/dirty_background_ratio thresholds to more aggressively get
> data onto disk..?

That idea had occoured to me too, but at least
vm.min_free_kbytes=4000000
vm.vfs_cache_pressure=200
did not prevent the problem from occouring.

Regards,
-- 
Michael Meier, Zentrale Systeme
Friedrich-Alexander-Universitaet Erlangen-Nuernberg
Regionales Rechenzentrum Erlangen
Martensstrasse 1, 91058 Erlangen, Germany
Tel.: +49 9131 85-28973, Fax: +49 9131 302941
michael.meier at fau.de
www.rrze.fau.de