[gentoo-user] Filesystem failure

[gentoo-user] Filesystem failure

[Willie Wong]

The root partition on my laptop just crapped out today. The kernel
logs report a problem with reiserfs. Reiserfsck suggests
--rebuild-tree, which AFAICT has a rather non-zero failure rate. So I
am getting myself mentally prepped for the perhaps necessary
re-install. 

I would appreciate greatly if people can offer me some advice/caveats
of any sort with regards to my plan. 

The run-down of the state of my laptop:

  1. Luckily/unluckily, I had separate partitions for /var and /home,
  neither of which seems to be affected (I need to still run smartctl
  and reiserfsck to verify that).
  2. It seems that the damage is limited to /usr (which is not mounted
  separately, so is a sub-directory of /)

The plan would be to bootstrap the system using a liveCD, I seem to
already have a workable /var (and hence an okay edb/cache and world
file), does it sound workable to just 
  a) get portage, system, toolchain running
  b) emerge -e world
?

Assuming my /etc isn't hosed, the system should still be bootable, if
I recover all the right files, right?

Thanks, 

Willie


-- 
Sortir en Pantoufles: up 104 days,  3:21
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Filesystem failure

[Willie Wong]

On Thu, Mar 22, 2007 at 01:08:24AM -0400, Penguin Lover Willie Wong squawked:
> The root partition on my laptop just crapped out today. The kernel
> logs report a problem with reiserfs. Reiserfsck suggests
> --rebuild-tree, which AFAICT has a rather non-zero failure rate. So I
> am getting myself mentally prepped for the perhaps necessary
> re-install. 
> 
> I would appreciate greatly if people can offer me some advice/caveats
> of any sort with regards to my plan. 

Let me be slightly more concrete. I plan to:

download a stage3 tar ball
unroll the tar ball on my ex-root partition
(follow the handbook)
chroot
emerge --sync
copy over my /etc from my backups
rebuild toolchain
mount my /var partition
emerge -e world

Is there any obvious flaws in that?

W
-- 
I have a speech impediment... my foot.
Sortir en Pantoufles: up 104 days,  3:55
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] Filesystem failure

[Mick]

[-- Attachment #1: Type: text/plain, Size: 1457 bytes --]

On Thursday 22 March 2007 05:37, Willie Wong wrote:
> On Thu, Mar 22, 2007 at 01:08:24AM -0400, Penguin Lover Willie Wong 
squawked:
> > The root partition on my laptop just crapped out today. The kernel
> > logs report a problem with reiserfs. Reiserfsck suggests
> > --rebuild-tree, which AFAICT has a rather non-zero failure rate. So I
> > am getting myself mentally prepped for the perhaps necessary
> > re-install.

You may find that a --fix-fixable does the trick and you don't have to 
re-install - this usually only works if you keep your fingers crossed at the 
same time.  ;-)

> > I would appreciate greatly if people can offer me some advice/caveats
> > of any sort with regards to my plan.
>
> Let me be slightly more concrete. I plan to:
>
> download a stage3 tar ball
> unroll the tar ball on my ex-root partition
> (follow the handbook)
> chroot
> emerge --sync
> copy over my /etc from my backups
> rebuild toolchain
> mount my /var partition
> emerge -e world
>
> Is there any obvious flaws in that?

The obvious flaw is that if the fs corruption is caused by hardware failure 
you may be back to square one in the near future.  OTH it may just be the 
result of fs fragmentation because you keep your /usr/portage on the same 
partition as /.  I guess you can try moving your /usr into a new partition 
and see how things evolve thereafter.  Did I say make a back up?

Good luck.
-- 
Regards,
Mick

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] Filesystem failure

[Willie Wong]

On Thu, Mar 22, 2007 at 08:42:35AM +0000, Penguin Lover Mick squawked:
> You may find that a --fix-fixable does the trick and you don't have to 
> re-install - this usually only works if you keep your fingers crossed at the 
> same time.  ;-)

I wasn't so lucky. :(

So now I am re-installing system. (Argh, I started an emerge -e system
at 4 in the morning, and I completely forgot to reset my /var/tmp to
the symlink that I usually used. This morning I woke up to see that it
stopped on the second package because I ran out of space on /var). 

> > Let me be slightly more concrete. I plan to:
> >
> > download a stage3 tar ball
> > unroll the tar ball on my ex-root partition
> > (follow the handbook)
> > chroot
> > emerge --sync
> > copy over my /etc from my backups
> > rebuild toolchain
> > mount my /var partition
> > emerge -e world
> >
> > Is there any obvious flaws in that?
> 
> The obvious flaw is that if the fs corruption is caused by hardware failure 
> you may be back to square one in the near future.  OTH it may just be the 
> result of fs fragmentation because you keep your /usr/portage on the same 
> partition as /.  I guess you can try moving your /usr into a new partition 
> and see how things evolve thereafter.  Did I say make a back up?
> 

Well, I hope not. I did run smartctl from the liveCD and the longtest
didn't see anything wrong with the disc. The damage was all very
localized to /. (in fact, /usr). 

It might have to do with a bump I gave the laptop: I was getting up to
use the bathroom and in bringing the laptop from my lap to the coffee
table, bumpped it on the armrest of my chair. I guess it could cause a
hardware failure, but so far I am praying it isn't. 

On the other hand, my desktop has survived several guttings due to bad
harddrives (once it was due to a flatmate who likes to slam his doors
shut... the shock travels across the plaster wall, down my desk, and
into my box), and I remember the symptoms of hardware failure to be
not so quick acting and to have given different error messages. 

As to backups, I always keep backups of the data partition. I figured
that whatever I can rebuild from the liveCD probably isn't worth the
backup. I'll see whether this experience teaches me otherwise. 

Ah... and one flaw in my plan that I discovered: my /etc configs
includes those for software I haven't yet rebuilt and those for stuff
in system that has been revised many times since 2006.1. For example,
after just copying over my backup /etc, I discovered I cannot log in.
The liveCD uses pam and I use shadow and some files were overwrote.
Good that I was logged in at another console (something I learned from
experience... that when fixing a computer never to log out completely
until you're sure you can get back in) and emerged shadow. Another
example is that portage complained about the overlays that exists in
my make.conf but not on my filesystem. 

W
-- 
If it weren't for the last minute, 
nothing would ever get done.
Sortir en Pantoufles: up 104 days, 11:57
-- 
gentoo-user@gentoo.org mailing list