On Wed, 14 Dec 2011 at 12:20pm, SINCOCK John wrote

Any advice on what might be going on here would be appreciated! - should
it really be necessary to delve into tuning vm system parameters just to
avoid memory fragmentation causing device drivers and system calls to
fail? To be messing around trying different vm paramaters just to make a
machine usable doesn't seem to make sense.

What I/O scheduler are you using? I've seen some suggest that when using a (real) RAID card, the deadline scheduler is preferred over the default cfq. Also, what FS are you using for your holding area? Finally, what is the nature of the data you're backing up (i.e. does it involve *lots* of small files)?

Example of memory fragmentation problems:

Nov 16 02:16:11 phoenix kernel: mount.nfs4: page allocation failure.
order:4, mode:0xd0

Out of curiosity, can you recreate the issue if you take NFS out of the equation?

Nov 16 02:16:11 phoenix kernel: Pid: 32280, comm: mount.nfs4 Tainted: G

What's tainting your kernel?

--
Joshua Baker-LePain
QB3 Shared Cluster Sysadmin
UCSF

Reply via email to