submonc_template.sb

#!/bin/bash
#SBATCH --job-name=add_jobname
#SBATCH --output=add_pbs_dir%x.o%j
#SBATCH --open-mode=append
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=128
#SBATCH --ntasks=add_np
#SBATCH --time=01:00:00
#SBATCH --partition=standard
#SBATCH --qos=taskfarm
#SBATCH --account=add_account


# Make sure any symbolic links are resolved to absolute path 
export SLURM_SUBMIT_DIR=$(readlink -f $SLURM_SUBMIT_DIR)

# Change to the directory that the job was submitted from 
cd $SLURM_SUBMIT_DIR

# MODULES
. env/add_modules
module load atp
export ATP_ENABLED=1
module list

# Recommended ARCHER2 settings that may help with performance at very high node counts
# Note, though that some problems have been observed when using these in testing,
#   including steadily increasing time to complete a timestep before failing (ucx).
#   https://docs.archer2.ac.uk/known-issues/
# module unload craype-network-ofi
# module unload cray-mpich
# module load craype-network-ucx
# module load cray-mpich-ucx
# module load libfabric
# export UCX_IB_REG_METHODS=direct  # (using ucx)

# Recommended ARCHER2 Settings, https://docs.archer2.ac.uk/known-issues/
export FI_MR_CACHE_MAX_COUNT=0   # (using libfabric)
# Recommended ARCHER2 I/O Settings, https://docs.archer2.ac.uk/user-guide/io/#recommended-archer2-io-settings
export FI_OFI_RXM_SAR_LIMIT=64K  # (using ofi for MPI)

export SUBMISSION_SCRIPT_NAME=add_scriptname
export MONC_EXEC=add_build_exec

export TESTCASE=add_testcase_mcf
export STDOUT_DIR=add_stdout_dirname
export CP_DIR=add_checkpoint_dirname
export RUN_NAME=add_testsuite_jobname
export MAX_CONTINUATION_RUNS=200

ulimit -c unlimited

echo -e "\nSubmission time: $(date)\n"
echo -e "Submitting from: $(pwd)\n"

. test_harness/continuation.sh
run_monc

# Output job statistics
echo -e "\nCompletion time: $(date)\n"
scontrol show job $SLURM_JOB_ID
sstat $SLURM_JOB_ID --format="AveRSS,MaxRSS"
# For finished jobs:
# sacct --format=JobID,AveCPU,AveRSS,MaxRSS,MaxRSSTask -j $SLURM_JOB_ID