/var corrupted.....

[Eric Schuele]

On 11/02/2006 10:21, illoai at gmail.com wrote:
> On 11/2/06, Eric Schuele <e.schuele at computer.org> wrote:
>> On 11/01/2006 11:05, Karol Kwiatkowski wrote:
>> > On 01/11/2006 17:40, Eric Schuele wrote:
>> >> Hello,
>> >>
>> >> [Running 6.2-PRERELEASE as of Oct 30th]
>> >>
>> >> My /var filesystem on my laptop died this morning.
> . . .
>> >> 2) If I have destroyed it what can I do at this point?  I have no full
>> >> backup of /var.  I had nothing of any real importance on there.  Some
>> >> MySQL data... but I've got that.  My package database comes to mind.
>> >> but nothing of any personal value... just stuff to keep the OS on its
>> >> feet.  So... if its gone... is there anyway to create a functional 
>> /var
>> >> filesystem that will allow me to "get back to work as usual"?  Or 
>> is my
>> >> only option a complete reinstall of everything?
> . . .
>> > The downside of this (option 2) is you'll loose some important
>> > information about your system, /var/db/pkg comes first to my mind.
>>
>> With respect to the package database...
>> I've seen plenty of threads from folks having lost theirs in some form
>> or fashion, and the solution always seems to be "reinstall everything".
>>   Well, ok... sounds like a PITA, but how hard can it really be.  I only
>> had 30-40 "apps" installed anyway.  With their deps it weighs in around
>> 350 ports total.  So I started to do just that.  Figured I'd reinstall
>> in the order I originally installed in the first place.  Starting with
>> Xorg.  I go to the port dir and `make install`, thinking it would
>> reinstall it and all its deps.  No go.  It does in fact reinstall Xorg,
>> but none of its deps because it finds them present.  Reinstalling 30-40
>> apps is one things, having to manually go in and do 350... now thats a 
>> PITA!
> 
> You might be able to force mount the dirty filesystem via
> mount -f
> You can also try
> dd if=/dev/ad0s1d of=some_dang_file_name*
> And then using mdconfig to play with the resulting file.
> mdconfig -t vnode -f some_dang_file_name -u 0 && \
> mount -f /dev/md0c /mnt (maybe?)
> If you can get the /var/db/pkg dir off nicely, good luck.

Thanks.  Good ideas.  I'll play with this when I have time.  But after 
using mtree to recreate the structure in /var (off the / filesystem), 
things came back online pretty well.  I'll most likely get things put 
back together and then just recreate the /var filesystem and copy 
everything back into it.

> 
> *(Note that this could take a long time on a 1 or 2G /var
> as it reads all of the empty blocks as well, you might want
> to hand it a bs= and a count= if you know about how much
> of /var was full at the time, man dd for more details. Also
> note that I have had faster results using sdd from ports)
> 
> I had a similar problem a while back and both methods
> were able to read some of the data from the former /var,
> however the /var/db/pkg directory was trashed and I ended
> up having to fall back on the "reinstall everything" method.
> My method ended up consisting of:
> 1) reinstalling portupgrade
> 2) reinstalling several high level programs (opera, mplayer,
> gnumeric, any window managers, & so on)
> 3) pkgdb -F which one at a time reinstalls everything depended
> on.  Make sure you have backups of any important files in
> /usr/local/etc as they may get overwritten.
> 

I'm presently doing this now.  I have reinstalled most, if not all, top 
level ports and will be running through `pkgdb -F` tomorrow.  I suspect 
that will take a fair amount of time.

Thanks.
-- 
Regards,
Eric