2

We have a warewulf cluster of about 20 nodes being standard PC boxes (I believe ATX, 4-year old) w/o hard drives - they boot from master machine via LAN. Some of the nodes tend to crash and we regularly have to go to the machine hall to press the restart buttons. We would like to avoid this. Is remote "hard" restarting nodes possible with standard ATX or we need some extra gadget for this functionality? The master machine doesn't crash - can we use its network card for sending the restart signal?

2 Answers2

5

If they're standard PCs then they don't have the functionality you require. However with something like a networked power socket (something like APC's AP7920) you can at least remotely control the power to the PC.

Cry Havok
  • 3,566
4

If there was a way to make computers respond to commands while they're not responding... we'd have machines that are always responding.

You probably want a lights-out kit, or a watchdog timer.

http://en.wikipedia.org/wiki/Out-of-band_management

http://en.wikipedia.org/wiki/Watchdog_timer

ta.speot.is
  • 14,323