xfs
[Top] [All Lists]

Re: restoring after running xfs_repair

To: Eric Sandeen <sandeen@xxxxxxx>
Subject: Re: restoring after running xfs_repair
From: Michael Sinz <msinz@xxxxxxxxx>
Date: Thu, 26 Sep 2002 15:24:49 -0400
Cc: James Rich <james@xxxxxxxxxxxxx>, XFS mailing list <linux-xfs@xxxxxxxxxxx>
References: <Pine.LNX.4.44.0209261002001.2090-100000@stumpy.chowhouse.com> <1033059609.7333.41.camel@stout.americas.sgi.com>
Sender: linux-xfs-bounce@xxxxxxxxxxx
User-agent: Mozilla/5.0 (X11; U; FreeBSD i386; en-US; rv:1.1b) Gecko/20020813
Eric Sandeen wrote:
On Thu, 2002-09-26 at 11:08, James Rich wrote:


I looked at what was created in lost+found and was easily able to put
things back in the right places.  Just the highest level directory
structure had been lost (i.e. the connections to /) so it was easy to just
mv 3145688 /etc.


Ok, good to know.


The original error would be most helpful, along with anything you can think
of that could have contributed to the problem...  (Any odd configuration,
patches, hardware, etc?)

No oops or anything. A complete description of the hardware is at www.supermicro.com. I did a shutdown -h now and on restarting it the machine paniced right after mounting /. The kernel is checked out from CVS as of a few days ago.

This sounds like the IDE-write cache does not flush before power off problem. There is a "fix" to this in some systems that has the shutdown process, after syncing/unmounting stuff do a powerdown of the IDE drives to force them to write their caches before doing a shutdown of the ATX power supply via APM/APCI.

You may wish to look into changing the write cache settings in the
drives (albeit they don't always listen to what you ask) or, even
better, to get your shutdown process to do the drive poweroff
(hdparm is your friend)

I think there is work going into the kernel to help with this but
most of shutdown is actually handled/managed from userland.

--
Michael Sinz -- Director, Systems Engineering -- Worldgate Communications
A master's secrets are only as good as
        the master's ability to explain them to others.



<Prev in Thread] Current Thread [Next in Thread>