Available CPU cores can be used by grid/user jobs according their contribution to build this cluster. User’s jobs automatic accounting group selection is based on linux group membership id -Gn and all other users that doesn’t belong to any experiment share one dedicated accounting group with fraction of available resources. List of currently defined HTCondor accounting groups
| Accounting group | Linux groups | Details |
|---|---|---|
| group_alice.user | alice | members of ALICE experiment |
| group_atlas.user | atlas | members of ATLAS experiment |
| group_auger.user | auger, cta | members of AUGER & CTA experiment |
| group_fermilab.user | dzero, dune, fermilab, nova | members of Fermilab experiments |
| group_user | — | other users that are not member of any experiment |
| group_hiprio | — | special (can’t be used without explicit approvement) |
Grid jobs use different accounting group (usually *.prod and *.test for special grid monitoring jobs) but they always share total amount of available resources for given experiment. Currently local user’s jobs can use up to 1/2 of resources dedicated to the experiment at the expense of grid jobs.
If you are member of multiple experiments and automatic group selection doesn’t give your jobs right accounting group, you can explicitly specify group in a job_submission_file
accounting_group = group_auger.user
English 
Čeština