! ! Settings for SLURM job manager on NOAA theia - for details and other ! options: "man sbatch" and ! https://rdhpcs-common-docs.rdhpcs.noaa.gov/wikis/rdhpcs-common-docs/doku.php?id=slurm#migration_to_slurm_workload_manager ! ! queue system (bsub,loadleveler) queue : slurm ! (1) passed directly (ie "as is" at the command line) !------------------------------------------------------------------ ! (--share or -s for example to share node(s) with other jobs) ! add here also the long options queue.slurm.submit.options : !queue.slurm.submit.options : --share ! Note that this is added to the 'submit.options' key value from the pycasso-tm5-expert.rc ! (2) Short options set by end user (and processed by submit_tm5_tools.py script !---------------------------------------------- ! First, list of queue options (same for each step): ! These each get prefixed with queue.slurm.options : job-name account time ntasks nodes output queue.slurm.option.output : %x.o%j ! job name: queue.slurm.option.job-name : ${job.name} ! NOAA-specific modification, appends jobstep number to job name; see ! TM5/base/branches/theia_slurm/bin/submit_tm5_tools__noaa_slurm.py queue.slurm.name.append.jobstep : T ! The "-A co2" option is specific to NOAA theia "co2" project queue.slurm.option.account : co2 ! queue.slurm.option.qos : batch ! Partition should be "batch" for parallels jobs and "shared" for ! serial jobs. Note that bigmem is also a valid option. The partition ! will be automatically set based on ntasks, unless it is set in the ! RC file. !queue.slurm.option.partition : shared !queue.slurm.option.partition : batch ! NOTE: o and e options are not active because they are not in the ! options list. They are here left as placeholders, showing use of ! . This was handled by special code (now commented out) in ! TM5/base/branches/theia_slurm/bin/submit_tm5_tools__noaa_slurm.py. ! That was for some other previous slurm platform/implementation, we ! are going to try using the default output file name. ! queue.slurm.option.o : ! queue.slurm.option.e : ! JOB TASKS, NODES, RUNTIME, AND MEMORY REQUIREMENTS ! ! Modify as appropriate for your particular run. ! For a serial job, put something like the following in your RC file. ! NOTE: no N (no. of nodes), and ntasks MUST be 1 !queue.slurm.option.ntasks : 1 !queue.slurm.option.time : 0:50:00 !queue.slurm.option.mem : 5g ! For a parallel MPI job, put something like the following in your RC file. ! NOTE: lack of "mem" option; not used for exclusive-node scheme for parallel jobs. !queue.slurm.option.ntasks : 48 ! this is the number of cores !queue.slurm.option.nodes : 4 ! this is the number of nodes (would result in 12 cores used per node) !queue.slurm.option.time : 4:30:00