easily reproducible filesystem crash on rebuilding array

[Emmanuel Florac]

Le Thu, 11 Dec 2014 12:39:36 +0100
Emmanuel Florac <eflorac at intellique.com> écrivait:

> What happens: while the array is rebuilding, simultaneously reading
> and writing is a sure way to break the filesystem and at times,
> corrupt data.
> 

I've rerun the same test (heavy read/write IO while rebuilding) with
disk drives write cache off (that is RAID controller is running
in write-back mode, but the independant disks caches are set to
write-through).

The filesystem went corrupted too, however much less than previously:
it went back up online after a umount/mount cycle. Nothing at all
appears in xfs_repair output. However the IO error is weird as the RAID
controller reported no such error.


Dec 12 00:40:18 TEST-ADAPTEC kernel: XFS (dm-0): xfs_do_force_shutdown(0x1) called from line 383 of file fs/xfs/xfs_trans_buf.c.  Return address = 0xffffffff8125cc90
Dec 12 00:40:31 TEST-ADAPTEC kernel: XFS (dm-0): xfs_log_force: error 5 returned.
Dec 12 00:41:02 TEST-ADAPTEC kernel: XFS (dm-0): xfs_log_force: error 5 returned.


Still investigating...

-- 
------------------------------------------------------------------------
Emmanuel Florac     |   Direction technique
                    |   Intellique
                    |	<eflorac at intellique.com>
                    |   +33 1 78 94 84 02
------------------------------------------------------------------------