Hi y'all,my lab is also active in the high energy physics community and we are sort of aliens there for our long standing use of Debian of almost 30 years. [1]
I've been participating in the Paris Meeting where Arne Wiebalck presented CERNs plans for a dual Linux ecosystem and among other things presented a short overview of our Debian roadmap and LTS and ELTS utilization. [2]
> so cvmfs official packages would be great.I explicitly asked about official Debian packages for CVMFS during a discussion at that meeting and the CVMFS developers quickly pointed out that they do produce Debian packages in a dedicated repo. My impression is that they would be happy to participate in building official packages and would first of all need some mentoring on how to do so, esp. getting there packaging in the right shape ie. linitan clean aso.)
>> It is worth mentioning that CERN's software group for scientific >> software (mainly ROOT, Geant4 and cvmfs) are an entirely different >> department than CERN IT. If they move more machines to Debian I doubt >> we will see upstream packaging as a resultGeant4 and ROOT have very diverse developer communities that not only transcend CERN IT but CERN as a whole. They also have potentially incompatible development and release models. For example we provide a manifold of versions of ROOT and Geant4 built with different compilers and different combinations of libraries to our users via "spack". This is something not really doable with official Debian packages, is it?
> And what about subcontracting the packaging ? The CERN can pour a few euros in this kind of task no ?
I doubt that anyone would fund this effort.More likely someone from the community will produce Debian packages (as it happened before eg. [3]). Maybe at this time, these efforts may become less ill-faited one-man-shows?
The reason I did not come forward earlier in this discussion is that we moved our HPC/HTC clusters from Debian to (Debian containers on) Rockylinux bare-metal servers some years ago, mainly to to better vendor support (Infiniband, Lustre etc.). As you may imagine that was also a very political decision, so an unbiased assessment of the pros and cons of Debian vs. RHEL derivative is difficult for me to make.
Best regards Christopher [1] https://www.debian.org/users/org/gsi.en.html [2] https://indico.cern.ch/event/1377701/contributions/5863767/ [3] https://snapshot.debian.org/package/root-system/ Am 28.05.24 um 09:52 schrieb PICCA Frederic-Emmanuel:
The Geant4 and ROOT work I did privately, and right now I don't actively work on those ITPs. At least on the ROOT I know they are interested, but they have other priorities given that ROOT is notoriously hard to package (huge monorepo with external deps, complicated build script, modified embedded clang, legacy names like "libCore.so" et al). I don't think Geant4 has any man-power to spare, and it also has their own packaging challenges (requires huge datasets to actually work).And what about subcontracting the packaging ? The CERN can pour a few euros in this kind of task no ?It is worth mentioning that CERN's software group for scientific software (mainly ROOT, Geant4 and cvmfs) are an entirely different department than CERN IT. If they move more machines to Debian I doubt we will see upstream packaging as a result, most likely they will just incorporate it better in cvmfs.so cvmfs official packages would be great. Cheers Fred
--
Christopher Huhn
Linux & web group
IT department
GSI Helmholtzzentrum fuer Schwerionenforschung GmbH
Planckstr. 1, 64291 Darmstadt, https://www.gsi.de/
Sitz der Gesellschaft / Registered Office: Darmstadt
Handelsregister / Commercial Register:
Amtsgericht Darmstadt, HRB 1528
Geschaeftsfuehrung / Managing Directors:
Professor Dr. Paolo Giubellino, Joerg Blaurock
Vorsitzender des GSI-Aufsichtsrates /
Chairman of the GSI Supervisory Board:
Ministerialdirigent Dr. Volkmar Dietz
Attachment:
smime.p7s
Description: Kryptografische S/MIME-Signatur