Summary: ODS crash

Matthias Kurz (mk@baerlap.north.de)
Fri, 18 Apr 1997 20:22:39 +0200 (MEST)

> I've lost two disks in a RAID5 volume. One was in Maintanence
> and one in 'Last Erred'. After i resynced the disk that was
> in Maintenance mode, the filesystem was corrupted and the
> machine constantly panicked with 'freeing already free inode'.
> Currently i'm resyncing the second disk, because after the
> first was finished i've got read errors and could not create
> at new filesystem.
> So my question: Is the raid really in a reliable state when
> the second resync finished ? Or must i reinitialize the whole
> raid ? If so - how can i stop the current sync.

Thanks to
Sun Support (sometimes i like mondays)
Thomas White <twhite@bear.com>
Jim Harmon <jharmon@telecnnct.com>

First a clarification: I did not really loose two disks. I lost
one disk. While the raid was working with one disk less, a second
disk stopped working. Reboot did not help, but after a power cycle
the second disk came back.

The answer to my question was: Yes, the raid itself will
be in a reliable state after the second sync, but the data on
the raid might be corrupted. In my case there was very much
data. I inspected only the most important things (needed for
the current work and/or not backed up) and found only few
corruptions. Inspection of the rest would have taken much time.
And there would have been the risk of overlooking something.
So i deleted the data and restored from backups. Can't say how
much of that was really corrupted.

In any case one should install the current ODS jumbo patch
(102580-xx). With this patch installed i probably could have
used the raid after the first sync was finished or even without
doing a second sync. The problem is, that while multiple components
in a raid are faulted you have only read access. ... see 1218867
in 102580-xx.README.
The filesystem was mounted read/write and that read only access
confused the system and led to the panics.

The other question was, whether there is a way to stop a sync.
The answer is: There is no command to stop a sync. But one
can shut down the system and 'unplug' the disk. This raises
an error - the sync is aborted and the disk is put back in
'maintenance'.

(mk)

-- 
Matthias Kurz; Fuldastr. 3; D-28199 Bremen; VOICE +49 421 53 600 47