RE: been a while before i could get back to this..

[Seth Mos <knuffie@xxxxxxxxx>]

At 17:59 19-2-2003 -0800, l.a walsh wrote:

> > Can you tell me what kernel you were running again? Was is
> > the 1.2 based
> > release with a bit older dump and restore?
>
> linux-2.4.20 (vanilla) with feb 04,04 xfs-all patch
>
> xfsdump 2.2.3.

Can you try a newer CVS perhaps. There has been some CVS breakage this 
month. You might be affected.

> No kernel panic or oops but it flies by awfully fast.  It's just
> so darn unfriendly...since I can't really do anything other than
> press reset at that point....Am in process of creating tbz2 archives
> for the other partitions right now...dang bzip2 is ..slow...I only
> have 6G of data, and its still cranking (backup ~3G so far)

if /var is seperate partition you might be able to find the filesystem 
shutdown message in the /var/log/messages log.

That might prove usefull.

> > From: Chris Wedgwood [mailto:cw@xxxxxxxx]
> > does xfsdump puke because it gets an error from the kernel, or does it
> > barf internally?
> ---
>         Hmm....it appears that something in the kernel driver thinks
> that the disk is corrupt so it unmounts the root drive.  xfs just
> sorta exits...  it flies by on the screen so fast and wasn't being
> cooperative when I wanted to scroll back.

As soon as XFS detects a corruption error it shuts down the partition to 
prevent worse. There must be something wedged that only gets triggered by 
xfsdump.

>         I wish there was a "pipe" mode where I could say, no, don't
> buffer, and flush it every character.  Inefficient as *bleep*, but
> in some situations....
>
>         Actually, now that I think about it, if I pipe the output
> to /tmp the pipe would already be open and its on a different partition,
> so theoretically, the output should be saved, no?

cd /tmp
nohup xfsdump -yada yada &
this should save at least the xfsdump output. Even if the root fs shutdown 
you should still be able to read and write other partitions.

>         Haven't tried strace...was hoping this bug would tickle someone's
> "aha" button and it was already 'solved'....

Nay.

>         I'll try to get more info ...but first need to rebuild the utils.
> Darn redhat RPM's aren't compatible with SuSE.  Maybe you could link it
> using the downrev of GLIBC and cover both RH and SuSE?  Just a thought.

You can rebuild the rpms safely on a SuSE machine.
You can try the following (not sure if it works) rpm --rebuild --nodeps 
xfsprogs-2.??.src.rpm

This should write some rpms to the rpm build directory, wherever that is on 
SuSE.

Cheers
--
Seth
It might just be your lucky day, if you only knew.