Hello
I want to use pp on both a largish cluster ( titan.uio.no ) and a "local workstations" cluster. But does there exist any scripts that starts ppseerver on my chosen set of machines, and shuts them down etc. over SSH?
Well, probably, but it's dependent on the cluster. I use PP on a PBS cluster and I start and stop the servers inside my PBS submission script. I imagine the same could be done for the other queuing software.
In my case, I use both cores on each node, so I have to remove the "doubling" from $PBS_NODEFILE:
# Run the ppservers on all the nodes.
mynodes=`cat $PBS_NODEFILE|uniq`
for i in $mynodes
do
ssh -f $i ppserver.py -a -p $portnum -w 2 -t 40 &
done
As you can see, I use -a to select auto-discovery, so I also have to use -p in case I don't want to step on other PP jobs running. I take care of killing off the servers with the timeout, -t. This way after 40 seconds of no calculation, the servers die and PBS can exit out.
Forum 
