Hi, Sometimes, when I'm running my simulation in the cluster and it gets hanged, the solver processes do not exit automatically and then I have to enter each machine, look for the processes PID's and kill them one by one. Does anyone have a Linux script for it or any other suggestion? Thanks a lot, Magnoli.

Check out the commands pgrep and pkill. Use at your own risk though, pkill can do serious damage when spread across a cluster. Test things out with pgrep first. To kill on a remote node:

rsh nodename pkill solver

You can do it in one line to all the slaves if your cluster has ganglia (look up online for further explanation) or the like; you would replace rsh with gsh for instance.

I repeat: be very, very careful with pkill.

-Alex

Thanks! I'll take care!