Re: [K12OSN] bad rpm database after restore from backup

[Les Mikesell <les futuresource com>]

On Sun, 2004-01-04 at 07:27, Julius Szelagiewicz wrote:

> 	no rpm operation worked before i blew away the database - that
> meant no yum as well. i cleaned out the rpm database just to be able to
> use apt to remove cups and install lprng. this din't go smothly, but it
> did go. still no joy with printtool, or printing, for that matter.
> 	the plan is: total destruction of th weekend by full install of
> 3.1.2 from scratch and than a very judicious restore:
> /home, /usr/local/bin, stop dhcpd, blow away leases files, in /etc:
> passwd, shadow, group, gshadow, dhcpd.conf, rsyncd.conf, ./init.d/nat ...
> 
> what am i forgetting here?

Just a lot of updates.  I'd copy all of /etc somewhere else so
you can pick individual files to copy back easier than doing
it from tape. Then do the apt-get or yum update to the fresh
install before dropping the old stuff back. 

If you have a similar 3.1.2 install somewhere, you might make
one last attempt to make rpm work by copying everything related
over from the other machine. 

If there is a lesson to be learned here it is that the best
way to to a major upgrade is to have a spare machine where
you can build it and cut over.  Then if there are problems
you can switch back quickly.  This isn't too expensive if
you have several servers because you only need to keep a
couple of spares so you can use one for testing and one
as the spare for repairs and rebuilds.  Hot-swap disk
carriers are a nice touch too, so you can just trade disks
instead of the whole machine.

And if it is any consolation, we have had even worse problems
doing upgrades on Windows servers.  Usually we make Norton
Ghost images of them for the restore because it is faster
and safer than most backups.

---
  Les Mikesell
   les futuresource com