Quote Originally Posted by aay
Quote Originally Posted by Henk Poley
What happens to a job if its node fails?:

...all jobs started from that node and all jobs which are currently running on that node due to migration are lost...
Deffinitely something to remember!!!
I guess that keeping the thread 'stopped' on the originating machine gives all kinds of (time-)dependency problems. But still...

btw, if you open your case you can just unplug the reset switch. And with an ATX case you might want to disable the '5s power-button pressed is reboot' in the BIOS. Anyways this could be a really bummer, you need to be sure all your nodes are rock stable. One whackling NIC contact brings down all machines, or at least the ones which sent out processes to this machine.

I don't know in howfar the Linux kernel supports the ATX specs. But you might be able to 'bend' the reset button to 'reboot'.

remember, users can be lusers...