[Gllug] Back-out plans on Debian/Redhat?

Khusro Jaleel kerneljack at gmail.com
Sat Mar 7 12:13:52 UTC 2009


Many thanks for your answers, much appreciated :-)

> Is your critical server tracking Debian stable?

It's simply tracking Etch but has not been dist-upgraded in a very
long time so there are a lot of updates due, including Apache, MySQL,
Kernel and Libc. AFAIK, Kernel and Libc are the ones which definitely
require a reboot. Otherwise I could have updated Apache, etc without
rebooting the server.

We are planning on using another spare server to serve the website
while updating and rebooting this one, of course.

>  * Test networks which allow you to duplicate the conditions of
>    production networks and then test the impact of upgrades

Yes, I'm thinking of learning some more about Cobbler/FAI/Puppet and
doing things like that from now on, but at the moment we do have a
spare server which is almost identical to the one we are upgrading so
we are going that route.

> So you would need to make sure you
> either had a repository available in which the previous version was the
> latest available, or just make sure you kept the old package and use
> dpkg to install it.

Yes, sorry what I meant was keeping an old copy of the .deb around,
which usually lives in /var/cache/apt/archives and using that to go
back. Since these updates are just Etch->Etch, nothing major should
change in Apache so I should be able to go back easily.

I've never read up on MondoBackup, will look into that, also that RAID
drive thing sounds like it should work. These servers are Dell
Poweredges with hardware RAID 1. Can you simply swap a secondary drive
to a primary in these systems?
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug




More information about the GLLUG mailing list