Re: how to make Debian less fragile (long and philosophical)

[goswin.brederlow@student.uni-tuebingen.de]

Justin Wells <justin@semiotek.com> writes:

> Boot with rescue disk is pathetic and abolutely unacceptable, as anyone
> who has ever remotely administered a machine will be itching to tell you.
> 
> I am not always there to stick the floppy in; but in 9 out of 10 failures
> I am left with some kind of usable shell.

I had major problems with libc6 on my m68k system, because when it
came out it wasn't compatible to the 68060 cpus. Installing it broke
everything.

Then I learned how to protect my system from such update failures. I
had a second system installed on another partition, with just the bare 
stuff. That system was mounted to /mnt/test and I did a chroot to it
to test the updates. That eigther failed or worked (and failed often
at first until the bugs where gone). After a failure, a "tar -xIf
testsystem.tar.bz2" restored everything and on I went. Once it worked
I installed the libc on the main system.

Another way to make sure you can recover from a broken update is to
have a second root partition (and don't tell me you don't have 30
spare Mb) and switch to that on failure. In case of an update, you
boot from the spare partition and update. If anything goes wrong,
reboot and your back to the old system.

May the Source be with you.
                        Goswin