Condor Submission to and from EOS

[j-s-ashley]

Submitting HTCondor jobs to and from EOS requires implementing a workaround. There are a few "how to" guides:

• https://batchdocs.web.cern.ch/local/file_xfer_plugin.html
• https://cern.service-now.com/service-portal?id=kb_article&n=KB0001998 (not Condor-specific, but potentially useful)
• https://batchdocs.web.cern.ch/tutorial/exercise11.html

[j-s-ashley]

The tutorial in the last link yielded an odd "directory does not exist error." I'm continuing to poke at this in the background while exploring other options.

[j-s-ashley]

The tutorial in the last link yielded an odd "directory does not exist error." I'm continuing to poke at this in the background while exploring other options.

I quadruple-checked the spelling, just to be sure, before anyone asks.

[j-s-ashley]

I managed to get Condor to save some output to EOS via this submission script. It requires that the executable be in the same AFS folder, and the direct output (sm_x_01.$(ClusterId).$(ProcId).out) still gets plopped into AFS, but the output generated by the executable script itself gets tossed to the output_destination, set to a subdirectory in EOS.

[j-s-ashley]

Turns out this issue has multiple parts: 1) MadGraph Generation 2) Analysis

I'll get to (2.) later. For now, (1.) seems to be doable by using a little-referenced series of switches within MadGraph's configuration files.

As an aside, set run_mode = 1, etc. only change the settings in the current instance of MadGraph. Insert the line (before "generate," to be safe, though this may not affect anything) "save options" to make the change persist through the docker instance.

Error popped up from first attempt. To be attached and reviewed.