Re: [PLUG] Weird APT error message

gabriel rosenkoetter <gr@eclipsed.net> · Sun, 17 Nov 2002 16:37:34 -0500

On Sun, Nov 17, 2002 at 03:44:24PM -0500, Mike Leone wrote:
> Haven't seen errors from any others. I'm thinking filesystem corruption,
> because I had to do a "reiserfsck --rebuild-tree" a number of times, until
> it finally finished. It kept "aborting".

It's definitely just the FS if /usr's on the same physical disk as
other partitions and you're seeing the other partitions fine. (I
mean, it's *possible* that only blocks within that partition have
gone bad, but when disks go, they don't usually go in the platters,
they usually go in the heads or cache, so you'd be seeing this all
over the place.)

For reference, what's the hardware concerned? (Bus, brand, brand of
the HBA, so forth...)

> Here's a weird addendum - I mounted a spare hard drive (also formatted as
> reiserfs), and did a "(cd /home/turgon && tar cf - .)|(cd /mnt && tar xvfp
> -)" - which, BTW, was something Gabe pointed out to me about 2 yrs ago, when I had
> other kinds of hard drive problems :-). Twice I got a kernel panic (I forgot
> to write down exactly where; "file-something.c", I think it was). This was
> on kernel 2.4.19. Needed a hard reset.

Yeah, but if the FS layer is giving the kernel broken information,
it's not shocking that you got a panic. You can easily get into
logically irrecoverable states that way. The kernel *could* just
bail on the transaction, kill the relevant processes, and go on with
life, but that wouldn't be very helpful for fixing the real problem.

Oh, and I prefer:

( cd /home/turgon && pax -r -w -v -pe . /mnt )

these days. ;^>

Unless you're after pure, unadulterated speed, in which case use
Schily tar. (Matters more if you're going to tape or to a disk media
with some non-standard block size. Schily's fast because he uses a
FIFO, mostly, which means that block size transformations are much
cheaper. He's got a good dd too. And that's Schily, as in the author
of cdrecord.)

> I booted into 2.4.16, and the copy at least finished, altho with errors. I'm
> going to try cleaning out my home directory first; then reformatting that
> drive as ext2, and trying again.

So, wait, this means that /home is blown too? Maybe you said it
already and I missed it, but what's the partitioning on the flaky
disk? Are *all* parititions actually showing errors?

> How can I do that from the rescue CD shell? I'm not visualizing this set of
> steps properly.

Can't help you there, since I haven't played with apt. (No Debian, and
I haven't taught alien/apt about NetBSD pkgsrc yet, though I'd like to
so that I can use just apt under NetBSD. Yes, I know about Debian
GNU/NetBSD, but it's not really what I want. I just want the package
management layer, not a fully-GNU userland.)

-- 
gabriel rosenkoetter
gr@eclipsed.net