[xfs-masters] [Bug 840] XFS internal error xfs_trans_cancel at line 1163 (debian lenny amd64)

[bugzilla-daemon at oss.sgi.com]

http://oss.sgi.com/bugzilla/show_bug.cgi?id=840





--- Comment #3 from Christoph Hellwig <hch at xfs.org>  2009-06-22 07:26:52 CST ---
I suspect the write back cache indeed means the writes only go to the
controller cache first and then go back to the disk, and the cache off means
the disk caches.  This is safe if you use the filesystem directly ontop of the
raid device, but not with device mapper / lvm inbetween as they can't pass
through the required write barriers in the kernel you use.  There should be a
warning every time you mount a filesystems that barriers are disabled because
of this.

For now I would strongly recommend to turn it off, although that has a
performance hit.  If you upgrade to 2.6.30 where we can send write barriers
through device mapper/lvm it should be safe to turn on again.

Btw, also make sure to run xfs_repair on the filesystem after the above
happened.

-- 
Configure bugmail: http://oss.sgi.com/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.