Question about EXT3 error messages in /var/log/messages

[Mark Cuss]

Ok - I did the fsck this morning, and there were definitely some problems on
the filesystem (ie - duplicate inodes, wrong reference counts, etc.) but
fsck seemed to clean them up fine...

This filesystem is on two disks configured in a striping RAID...  These
disks were originally in my old server, but I plugged them into my new SCSI
disk array and the system recognized the md device and everything seemed
OK - I guess I should've known to do an fsck before putting the system into
service...

Thanks again
Mark

----- Original Message ----- 
From: "Stephen C. Tweedie" <sct at redhat.com>
To: <mcuss at cdlsystems.com>
Cc: <mbasil at alabanza.com>; "ext3 users list" <ext3-users at redhat.com>
Sent: Thursday, April 15, 2004 5:00 AM
Subject: Re: Question about EXT3 error messages in /var/log/messages


> Hi,
>
> On Wed, 2004-04-14 at 20:10, Mark Cuss wrote:
> > Okay - so it is the major and minor numbers - thanks!  That means that
md2
> > is the culprit...
> >
> > Does this mean that I have a drive failing in this raid or could the
> > filesystem just need an fsck?
>
> A drive failing should show up as IO errors in the logs, and the md
> layer automatically switches out drives which give errors.  So it's not
> a drive failing in the usual sense.
>
> You've just got corrupt metadata on disk.  How it got there is pure
> speculation --- the disk, controller, memory, CPU or software might be
> at fault, and it's impossible to tell at this point.  But a fsck is
> definitely recommended, as some types of on-disk corruption can spread,
> corrupting other data as time goes on (in particular, if an indirect
> block or bitmap gets corrupted then disk blocks belonging to one file
> can get overwritten by being reallocated to another file.)  You don't
> want to wait for that to happen!
>
> Cheers,
>  Stephen
>
>
>