Hi,
Yesterday I rebooted a server with a recent XFS CVS (downloaded/compiled
Tuesday of this week; the kernel reports at boot time - 'SGI XFS
CVS-10/08/02:05 with quota, no debug enabled'. Since then the kernel has
shut down the one XFS filesystem on this server:
Oct 10 14:22:10 blue01 kernel: xfs_force_shutdown(md(9,0),0x8) called
from line 1041 of file xfs_trans.c. Return address = 0xc01d9138
Oct 10 14:22:10 blue01 kernel: Corruption of in-memory data detected.
Shutting down filesystem: md(9,0)
Oct 10 14:22:10 blue01 kernel: Please umount the filesystem, and rectify
the problem(s)
I've rebooted, ran xfs_check (and xfs_repair, which was necessary on the
first occasion, the 2nd time - above the 'xfs_check' was clean) and
remounted the filesystem OK.
Is there a possible problem with this CVS that would explain this, I was
running (and have now reverted back to a XFS CVS kernel a few weeks old
(SGI XFS CVS-09/15/02:17 with quota, no debug enabled). I've not seen
the 'xfs_force_shutdown' problem before but have had repeated system
panics/lock ups. - though I guess it maybe that recent CVS is better
able to trap errors within the XFS code?
The server is a NFS server with no user code running directly on it. The
filesystem in question is a RAID 0 'md' stripe across 2 HW RAID5 units
(~1 Tbyte in size).
Regards and thanks for any suggestions.
Ian Hardy
Information Systems Services
Southampton University
UK
|