[Gllug] best way to update a single production server?

[James Courtier-Dutton]

2009/5/1 Khusro Jaleel <kerneljack at gmail.com>:
> Hi everyone,
>
> We have a very important production server that needs updates (just
> simple Debian Etch updates, not Etch to Lenny). There are several
> updates, including kernel and other updates, etc.
>
> Since unfortunately we have only this server, and it MUST be up 24/7,
> I'm not sure how to apply these updates, since a reboot is needed.
>

I have never heard of a setup where an individual x86 server MUST be
up 24/7. That is just plain impossible. Even five nines telco highly
reliable switches cannot be up 24/7 and no-one tries to make them. The
requirement is normally that the service MUST be up 24/7 with some
reliability figure attached. E.g. 99.999% of time.
One then has maintenance windows, where one has a chance to do
maintenance on a particular server. During this maintenance window,
the server is off-line, with other servers taking over the load. One
is free to reboot the machine as many times as one likes without
impacting the service.
To achieve your aims, you have to create this maintenance window.
The most common way to achieve this is to have at least two servers.
While one is active, the other may have a maintenance window.
Another way to achieve this is to relax the 24/7 requirement and just
plan maintenance windows at quiet times.
Creating maintenance windows is often simplified if the service is
contained in a Virtual Machine.
When one moves a virtual machine from one physical machine to another,
it keeps the same IP address. When the move happens, the virtual
machine sends out a special packet to inform the network of the move,
so packets get re-routed correctly. For those interested, the special
packet is called a "Gratuitous ARP".

Cheers

James
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug