Dear all,
I have been able to run some FCC analysis code locally, but when I try to run it on HT condor I get some unusual errors.
I have pasted them below…
Does anyone know how I could solve this?
Best wishes,
Sarah Williams
In case you’re interested in the error… I am getting this error in the HT condor error log:
You are probably running some locally compiled code (from the LLP fork of FCCAnalyses ?) when you run locally. While when you run with the batch option, the standard FCCAnalyses library is used maybe (I’m not fully sure here, but that would explain what you see).
The code that you are using (with FCCAnalysesV0) has been merged into the central FCCAnalyses repository yesterday only. And I see that /cvmfs/fcc.cern.ch/sw/latest/setup.sh
takes a release from 2023-01-15. Hence the central FCCAnalyses does not know indeed about FCCAnalysesV0.
If you are indeed using the central library when you run on the batch, I think you just have to wait for a new key4hep stack to be built and installed, that takes the last version of FCCAnalyses (assuming that the local fork did not contain any other functionality not integrated in the central repository).
If that is not the reason of the problem: a full A-to-Z recipe to reproduce the problem will certainly be needed for someone to take a look.
Thanks for your post. Can you confirm that you correctly source the setup.sh from your local FCCAnalyses build at the beginning of the shell script which is ran on condor nodes?
Hi all,
Thanks for the input here- this is interesting and seems consistent…
Do you know if there’s a way to get around this for now?
Let me discuss if I can get others to check before coming back to you…
Essentially I am running instructions here:
But get stuck running analysis_stage1.py on HT condor.
Any pointers would be greatly appreciated.
Sarah
Hi @brfranco and @eperez ,
I am just coming back to this as I believe @jalimena and Yuliia Borysenkova have been experiencing the same problem.
I checked and my local version of setup.sh points to the same release.
Do any of you have any idea either how quickly the next version will be built, or if there’s a workaround here?
Best wishes,
I came across a similar issue, and one quick fix that I found was to source my local setup script in config/FCCAnalysisRun.py instead of the default /cvmfs/sw.hsf.org/key4hep/setup.sh. This is probably not the best “long-term” solution, but maybe it can help in your case too.
Hi @jsmiesko and @lportale ,
Thanks for the input here, and sorry this is likely going to be a stupid question. My local setup script is still pointing to the 15th January one…
can I check what I should do with the PR below? Should I be pulling it into my area? (and potentially stupid question… how?)
Help getting this working is much appreciated, as its becoming increasingly time-consuming not being able to run things on condor…
Sarah
Hi all, @jalimena very kindly merged these changes into the FCC-LLP repo that we were running off, and I have now been able to run on HTcondor without the same error as before.
It seems to have ran ok, but I am still getting some errors…
File “/usr/sbin/make-jobfeatures”, line 77
os.umask(0022)
^
SyntaxError: leading zeros in decimal integer literals are not permitted; use an 0o prefix for octal integers
If you want to look at the full outputs they are here:
/afs/cern.ch/work/w/williams/public/FCC-LLP-tutorial-Winter23/FCCAnalyses/BatchOutputs/2023-03-21_14-24-03/p8_ee_Zee_ecm91/
Thanks for any input,
Sarah