fallback-reboot

fallback-reboot is a last resort, when you have a wedged system that you'd
like to reboot remotely. It requires that the machine be pingable, but it
does not require that machine's hard disk be functional, nor does it
require fork() or exec*() to be working.

fallback-reboot reboots the machine hard - it makes no attempt to sync
your disks. That's because fallback-reboot assumes that you've already
tried all the methods of doing an orderly shutdown you can think of, and
fallback-reboot is all that's left to try, short of hoofing over to the
machine or calling someone on the phone. In theory at least, it should not
require paging or swapping to be working either.

fallback-reboot currently only supports Linux and Solaris, but adding
support for other operating systems would most likely be fairly trivial.

http://dcs.nac.uci.edu/~strombrg/fallback-reboot/

I recently added a challenge-response authentication mechanism to it, to
deter replay attacks. You can use telnet as a client still, but there is
not a "fallback-reboot-client" program written in python included in the
distribution.

The daemon itself is in C, of course, since it requires minimal
dependencies.

Enjoy.