Aha!
We do periodically run out of memory on our ops, and this certainly
could be it.
Nice detective work!
Thus spake Mike Ryan on Thu, Dec 03, 2009 at 02:02:05PM -0800:
tb-admins:
The DETER team recently uncovered a bug in FreeBSD's NFS code that
causes the kernel to leak memory. If left untreated this bug can exhaust
kernel memory, causing a panic. As far as we can tell, this affects all
versions of FreeBSD currently in production.
We submitted a patch which has been pushed into FreeBSD CVS. The file
affected is src/sys/nfsserver/nfs_serv.c:
http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/nfsserver/nfs_serv.c?sortby=date
An NFS client which tries to delete an inexistent file will trigger the
bug. This can result from rebooting the ops node while experiments are
running.
To test if you've been affected, on your ops node run:
$ nfsstat | grep -A1 faults
Server Faults
0
If this number is greater than zero (or worse, climbing), then your NFS
server has been leaking memory. Regardless, we recommend patching your
systems.
For more details about the problem, see the original PR:
http://www.freebsd.org/cgi/query-pr.cgi?pr=140853
Mike Ryan
_______________________________________________
Testbed-admins mailing list
Testbed-admins@flux.utah.edu
http://www.flux.utah.edu/mailman/listinfo/testbed-admins
_______________________________________________
Testbed-admins mailing list
Testbed-admins@flux.utah.edu
http://www.flux.utah.edu/mailman/listinfo/testbed-admins