Cluster mode

Asked by Sebastian Urrutia-Quiroga on 2019-06-23

HI all,

I'm trying to run MG5 in my university's PBS cluster, but I'm having some questions about how to implement it. Specifically, I would really appreciate if you can address me about the use of the following options:

- cluster_size
- cluster_temp_path
- cluster_local_path
- cluster_status_update

I read the description in the configuration file, but I still don't understand. Thank you so much in advance,

Sebastian

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Olivier Mattelaer
Solved:
2019-06-27
Last query:
2019-06-27
Last reply:
2019-06-26

Hi,

- cluster_size

This is used only by loop_induced cluster in order to split the job close to that size of jobs.
No effect for the rests.

- cluster_temp_path

This a path where the cluster will cp all the files before starting the run. You can typically use that for using some local scratch.

- cluster_local_path

This is designed in order to avoid to have to copy your pdf set to each node if you use the above option (or to avoid that the node read them via the network).
It is designed to link to a cvmfs space where you can find the pdf sets.

- cluster_status_update

This control how times the qsub command will be run in order to check the status of the submitted job.
this takes two integer, since we have a short/long waiting time between update to try to not surcharge the pbs server with million of request but still be quite reactive for small jobs.

Cheers,

Olivier

Thank you so much!

Thanks Olivier Mattelaer, that solved my question.