If you specify a node_list in moab, the resulting msub command does not use the num_nodes options.
If you specify a node_list in slurm, the resulting sbatch command contains the num_nodes and the node_list items. In the testing branch I have made slurm not specify the num_nodes. Is this going to be a problem for LLNL or someone else?
The problem that still exists with both implementations now is that the PV_NNODES and PV_NPES environmental variables do not have the correct values if the node_list is a different size than the num_nodes option. I would like to make PV_NNODES and PV_NPES use values based of the node_list if it exists and completely ignore the num_nodes option.
LANL CCS has requested that I add support for lists in the node_list option. I would like to go this route so that you specify different size lists in the node_list option and still have reasonable behavior. Does anyone object to this modification?
If you specify a node_list in moab, the resulting msub command does not use the num_nodes options.
If you specify a node_list in slurm, the resulting sbatch command contains the num_nodes and the node_list items. In the testing branch I have made slurm not specify the num_nodes. Is this going to be a problem for LLNL or someone else?
The problem that still exists with both implementations now is that the PV_NNODES and PV_NPES environmental variables do not have the correct values if the node_list is a different size than the num_nodes option. I would like to make PV_NNODES and PV_NPES use values based of the node_list if it exists and completely ignore the num_nodes option.
LANL CCS has requested that I add support for lists in the node_list option. I would like to go this route so that you specify different size lists in the node_list option and still have reasonable behavior. Does anyone object to this modification?