Why to have a maintenance partition
Ross wrote:
Love ya Rick, but you make things too complex.
In this case, it's all about downtime.
When the debacle with /usr/lib happened, I had a big problem: The machine is a 2U rackmount box with no CD-ROM drive, and all I had to recover with was floppy disks. E.g., take your pick: Tom's Root/Boot, Slackware installation floppies, Debian installation floppies. So, I think I booted one of the former, to try to figure out what had happened (which took a while in itself).
When I finally did figure out that /usr/lib's contents had gotten massively mislinked, rebuilding the libs from the floppy-based maintenance system alone proved impossible, because there were just too many incompatibilities between the software packaged by Tom Oehser on the floppy and the HD-based facilities I was trying to repair. I couldn't even get dpkg and apt-get to work.
In the end, I was obliged to do things the hard way: I copied the package database (/var/lib/dpkg/status) over to another machine (along with /etc/* and other essentials) blew away the entire machine contents except for /home, /usr/local, and parts of /var, did a fresh Debian-base install, reset apt selections from the preserved package database, did apt-get fetches to get the installed packages back, copied the system configuration back, and was back in business.
But that took too damned long and too much trouble. Having a maintenance partition, ready to use and configured close enough to the in-service one, would have made the whole affair tremendously easier, so that's when I decided to have one around on critical machines from that point forward.
Rick Moen
rick@linuxmafia.com
If you lived here, you'd be $HOME already.