ReFrame tests

Running weekly tests for production

Note: performance logs are only sent to syslog/ELK when run as user vsc10001.

# login to Hydra as vsc10001
~/reframe-tests/run_weekly.sh

Running tests in your account

git clone https://github.com/vub-hpc/reframe-tests.git
reframe-tests/run.sh <options>

Examples

# all Lmod tests as jobs in compute nodes
reframe-tests/run.sh -c lmod
# Lmod test LmodTestJavaMemory in the local node
reframe-tests/run.sh -c lmod -n LmodTestJavaMemory --system local
# OSU tests compiled with foss/2022a toolchain
reframe-tests/run.sh -c osu --valid_prog_environs foss-2022a
# OSU tests in ReFrame partition skylake-mn-mpi-ib
reframe-tests/run.sh -c osu --partitions skylake-mn-mpi-ib
# GROMACS GPU test in ReFrame partition zen2-ampere-sn-gpu
reframe-tests/run.sh -c gromacs_bench --partition zen2-ampere-sn-gpu -n GMXBenchMEMSingleNodeGPU
# Slurm tests as jobs in compute nodes
reframe-tests/run.sh -c slurm

Location of ouput and log files

All output is written to VSC_SCRATCH_VO_USER/hpc-reframe-tests/

• logs/ ReFrame log files
• perflogs/: performance logs
• stage/: build and run scripts, (job) output, and (job) error files

Using old modules

In VUB-HPC clusters, Lmod emits a warning message (and returns non-zero exit code) when loading an old module, which breaks the ReFrame tests. To run suppress the warning and non-zero exit code for old modules, run the tests as follows:

REFRAME_QUIET_MODULE_LOAD=yes reframe-tests/run.sh <options>