Freinds,
Please recall that a week before I put up a similar problem and I was
told that the problem could be beause of the ancient code of XFS (2
years old) I am running on my box.
Answer was that new XFS code is stable, proven and works without any
issues and there no *VERY CRITICAL* issues like data corruption.
Aman's problem's source may have been different to mine but the
bottomlime is that "There could still be a major stability issue even
with newer XFS codes".
Steve/Lonnie : I don't want to sound un-neccessarily finiky and I
completely trust communitiy's ability to set things right, my only
submission is that there could still be grey areas
Best Regards
Mahesh
______________________________ Reply Separator _________________________________
Subject: RE: Data Corruption Problem
Author: ashahi (ashahi@xxxxxxxxxxx) at internet
Date: 7/30/2003 6:39 PM
Hi Greg,
I have a shared storage solution as you mentioned, and am using our
own HA software. We do take care of the fact that the current LVM is
not cluster aware.
regards,
Aman.
-----Original Message-----
From: Greg Freemyer [mailto:freemyer-ml@xxxxxxxxxxxxxxxxx]
Sent: Wednesday, July 30, 2003 12:14 AM
To: Aman Shahi
Cc: 'linux-xfs@xxxxxxxxxxx'
Subject: Re: Data Corruption Problem
Are you saying you have a shared storage solution with 2 LVs that you
move back and forth between 2 clustered nodes?
If so, I personally would use Linux-HA (heartbeat) in combination with
EVMS 2.0.
EVMS 2.0 is an alternative to LVM, but it understands clustering and in
particular it can use heartbeat for its cluster comm.
If you use LVM, I think you have to add a lot of custom scripting to
make sure LVM does not screw up due to clustering issues. (i.e. LVM is
totally cluster unaware. EVMS is cluster aware.)
HTH
Greg
--
Greg Freemyer
On Tue, 2003-07-29 at 17:45, Aman Shahi wrote:
> Hi,
> I am using linux 2.4.20 + LVM 1.0.7 +
> XFS(snapshot-xfs-2.4.20-2003-04-07_05:19_UTC with ACLs, no debug enabled).
>
> I created couple of Logical Volumes using LVM, and then created/mounted
file
> system over it. I am running some NFS Client doing I/O over different
files
> in these file systems. I am doing Failover/Failback testing. That is I
have
> one filestem attached to one node and other to the second node. When I
fail
> one of the node, the other node takes over the file system of the second
> node. When trying to mount the file system of the second node, I am
getting
> File System corruption.
>
> Could anybody tell what is the problem here. Attached here is the output
> from "dmesg".
>
> thanks in Advance,
>
> Aman.
|