In Development, do not use or at your own risks...
This library triggers a slurm job based on input file and a .hero matching file.
The goal is to use this library in a file upload hook mechanism to trigger a job on uploaded files, a serverless data service.
.hero file can be located in user home directory or a subdirectory of the file. If multiple files exist, the nearer of input file will be used.
Example:
hero: # map of experiment and job definition
exp1:
methods: [] # [optional] list of methods to apply rules [add, delete, edit]
rules: # list of path (regexp) to match against input file to run a script
- /home # example all files under /home
cpus: 3 # [optional] number of cpus
mem: 10 # [optional] optional memory requirements in G
time: 05:00:00 # [optional] slurm time
queue: long # [optional] slurm partition
scripts: # list of commands to execute with {{.File}} referencing input file
- ls -l {{.File}}
- md5sum {{.File}}
Job execution is done on behalf of selected user with slurm --uid/--gid options, meaning generated script must be executed as root (sbatch xxx).
Apache-2.0
2021 Olivier Sallou olivier.sallou@irisa.fr