Reputation: 20774
The cluster in my campus uses PBS TORQUE for job management.
There are a couple of nodes that are out of date in software. If my job gets sent to one of those nodes, it will fail. Is there a way to exclude a specific node (or list of nodes) in a job request?
Upvotes: 2
Views: 2073
Reputation: 745
I'm not sure you can specify a set that your job can choose from. You can specify a list of nodes but you will have to wait for all of them to become free before your job will run. Depending on the number of nodes you need to exclude you can do something like this.
cat $PBS_NODEFILE | grep -v badnode1 |grep -v badnode2 > goodnodes
NP=wc -l goodnodes #need back ticks around "wc -l goodnodes"
mpirun -np $NP -machinefile goodnodes ./a.out
This isn't ideal if you are in any way charged for resources but it should help you get your jobs run.
Upvotes: 0