On Tue, 2010-10-26 at 11:31 +0200, Peter Hall wrote:
> after a power failure I have severe corruption on a /home
> JFS-partition of 2TB.
> jfs_fsck fails with a lot of error-messages, ending in
> "Duplicate block references have been detected in Metadata. CANNOT
> CONTINUE. [fsckmeta.c:349]"
> I have a 1MB log-file of "jfs_fsck -d" that I can attach, meanwhile I
> have attached an edited version of that logfile.
> I am able to mount the partition read-only and have been able to
> salvage about 290GB of the 900GB of data in total. I have some backups
> but would be VERY happy if I could reclaim most of the 900GB instead.
Even if we were to hack jfs_fsck enough to get past the duplicate
references, I doubt it would recover anything you can't get by mounting
read-only. When it finds problems, jfs_fsck usually tosses away
whatever is corrupt, and recovers what's left. The read-only mount is
most likely seeing everything that is recoverable.
> I have not yet run "jfs_sck --omit-journal-replay" but I am unsure if
> this would help at all.
It won't. That only helps when jfs_fsck aborts during log replay (phase
0). Since it continued on its own, --omit-journal-replay won't make any
> I have searched the list and googled but have not found anything solid
> to go on. Any advice would be helpful.
I think you're best off recovering from your backups and whatever you
were able to save from the read-only mount.
> Thanks in advance,
IBM Linux Technology Center