[K12OSN] I need opinions....backup idea for when Terminal server down and I'm not around

[Les Mikesell]

On Sun, 2005-07-10 at 08:26, Jim McQuillan wrote:
> >    * The most likely server failure is a disk drive - use software raid
> >      disk and hot-plug disk drives if possible
> >    * The next most likely server failure is a power supply - servers
> >      with dual redundant power supplies are nice, or at least with a
> >      hot-pluggable power supply if not redundant.
> 
> Hmm,  hot-pluggable, but not redundant?  /me wonders what's the point.
> if a non-redundant power supply fails, everything goes cold  :)

It's one thing to go down, it's something else to be down a week while
you get replacement parts.  My favorite approach to minimizing downtime
is to have plenty of spare parts on hand, generally in the form of a
spare server that can be used for testing new software but can be put
into service or have parts pulled in an emergency.  You need the test
box anyway and save money by making it the same as the backup - you just
have to be careful never to do anything on it that you can't shut down
at a moment's notice. 

In a k12ltsp environment with more than one server, it makes sense to
put the home directories and authentication service on a heavy-duty
server with mirrored swappable drives.  Add dual power supplies and a
good UPS and you've covered most of the possible hardware failures.  If
you keep a spare server for this, in the unlikely event that something
other than a drive or power supply fails, you can pull the drives from
the production box, put them in the spare and be back up immediately.
You still need backups, but this makes it a lot less likely that you
will need to take the time to restore them.

With your local files all on one server, the others are more or less
disposable and interchangeable so you can use cheap commodity stuff
and just swap in a spare if it breaks, almost like the terminals.
The one thing I think is worth spending extra for is the swappable
drives, especially if you have several servers spread over some
distance because it means you can do upgrades or repairs by cloning
a disk from your test box and have someone else install it with just
the downtime it takes to reboot.  

-- 
  Les Mikesell
   les at futuresource.com