This project has utilities meant to be used in IHEP. They are site-specific and not useful for anyone outside IHEP.
One can submit jobs to the cluster by doing:
job_run_commands -f jobs.txt -e my_env
where jobs.txt
is a file with the commands to be ran, one per job, e.g.:
run_calculation -i 0
run_calculation -i 1
run_calculation -i 2
run_calculation -i 3
would run in parallel run_calculation
for different arguments. The flag -e
specifies the virtual environment where this needs to be executed.
To modify the amount of memory used by the held jobs and release them do:
jmanager -m 8000 -t 0
-t 1
will do a test run.
To submit arrays of jobs to HTCondor in IHEP do:
from ihep_utilities import JobSubmitter
d_job = {
'j1' : ['echo 1', 'echo 2', 'echo 3'],
'j2' : ['echo a', 'echo b', 'echo c'],
}
sbt = JobSubmitter(jobs=d_job, environment='None')
sbt.run(skip_submit=False)
Where
- The
j1
,j2
, etc can be replaced with the user's commands environment
is the user's conda environment where this will runskip_submit
can beTrue
for dry runs.
The outputs will go to:
/publicfs/ucas/user/campoverde/Jobs/run_commands
but this can be changed with JobsSubmitter._log_root