The anthill cluster currently uses the old Sun Grid Engine (SGE) as its scheduler. However, since Oracle bought Sun they have made it a closed source product, and so we will likely be switching to the Open Grid Scheduler (http://gridscheduler.sourceforge.net/). OGS is a fork of SGE, and has many features in common. Here is a website that has a lot of details about SGE and will be updated towards OGS in the future: http://arc.liv.ac.uk/SGE/howto/. In the read more I provide some examples of common commands that we use with the SGE.
-l h=node1 use only node 1
-pe make 8 use all 8 cores on a single node. This will make the job take over the whole node
-S /bin/bash ensure that SGE uses the bash shell. The default is tcsh
qmod -c all.q@nodeX clear stuck nodes
If you want to use a specific node, use this command. On anthill node1 has a lot more memory (24,748,432k) than the other nodes, and so you can use this command to request that specific node.
-l h=node1
To use all 8 cores on a single node use this command. This will make the job take over the whole node
-pe make 8
To ensure that SGE uses the bash shell (the default is tcsh). If you have more complex scripts, they will fail unless you use this!
-S /bin/bash