Hello all,
There are issues with the XFS support in the older (not cutting-edge)
SL5 kernels, which can cause xfs filesystems with more than a couple
of T of data in them to become unstable. No data corruption is caused,
the issue being related to spurious and false error signals from the
xfs kernel module.
In order to improve stability, I suggest that everyone with SL5 pool
nodes who isn't on kernel 2.6.8-164.15.1.el5 or higher and is using
xfs should upgrade to that kernel.
You'll need the 64bit
kernel-module-xfs-2.6.18-164.15.1.el5-0.4-2.sl5.x86_64, which appears
not to be in the CERN SL5 repository. We found it was available from
IN2P3's repo at:
http://quattorsrv.lal.in2p3.fr/packages/os/sl540-x86_64/errata/kernel-module-xfs-2.6.18-164.15.1.el5-0.4-2.sl5.x86_64.rpm
This should be (at least, it was for us), a simple upgrade.
(Thanks to Elena, by the way, for noticing and submitting this kernel
recommendation.)
Sam
|