[plug] ups driven automatic server shutdown

Adam Hewitt ahewitt at theozhewitts.com
Fri Jun 20 16:59:28 WST 2008

Harry McNally wrote:
> Hi Adam
> A few people have said nut but I'm not sure that it meets your requirements.
> Servers running nut can "listen" to a common server over the network. The
> common server is generally the one connected (via serial, USB, or simple
> digital alarm signals) to a common UPS that powers all of the servers. On
> receipt of an alarm from the common server (over the network) the "slave"
> servers can autonomously shutdown (and you could arrange the shutdown scripts
> to bring down services gracefully on that server).
> But your application describes the reverse. Rather than autonomous shutdown of
> the slaves, you want a central server to supervise the shutdown as a sequence.
> While the alarms for each "slave" server could occur at timed intervals (nut
> scripts allow this), this would rely on the shutdown of each server occurring
> in a maximum time window. If you can guarantee that ..
> Have a look over nut anyway. It has quite comprehensive facilities and I may
> not have dug deep enough when I looked at what it can do. You may also be able
> to extend what it does to minimise scripting to achieve what you need. If the
> nut running on slaves can reply when all services that matter are closed, then
> this might permit a VerySmallScript on the "master" server.
> If 140 servers are running off a common UPS, you have a stout UPS there :P
> Maybe I'm assuming the architecture incorrectly.
> All the best
> Harry
> ps Top post be damned!

The 140 servers are actually VM's running inside a blade center and a 
few other random servers....but yes the UPS's are rated up to 140kVa 
each, and we have two in parallel, so they are quite beefcake.

Thanks for all your comments, looks like I will be giving NUT a deeper look.


More information about the plug mailing list