AFS phaseout coordinator meeting

Europe/Zurich
513/1-024 (CERN)

513/1-024

CERN

50
Show room on map
Jan Iven (CERN)
    • 16:00 16:20
      AFS Status update 20m
      Speaker: Jan Iven (CERN)
      AFS status update:
      Q (Elena G.): When will the web pages be migrated?
      A: Not yet fully defined; webservices can already serve files from EOS; we have asked the owners to do the migration manually; there might be also old websites that might not need migration (to be confirmed by their owners)
      Comment (Joel): ACLs are different and some dynamic websites are a pain to migrate; no access to the webservices logs
      C (Andreas W): there should be logs available; sometimes it's not obvious if the errors are related to the websites
      C (Jan): We still need action on the users' side in order to manually point their websites to the new location
      C (Natalia R, CMS): Do we expect for the AFS area(s) to be available until CMS moves to Rucio (as we would like to converge to using only one system)?
      C (Jan): We want to be finished with the migration before Run3. Please contact the AFS migration coordinators for CMS (A Pfeiffer, S Malik Muzaffar) for more clarifications; If Rucio migration is fast (couple of months), then probably you do not need to worry; if it's going to take more (years), then please consider moving; please check your timelines and send an email/ticket for this
      Q (Giovanni, BE-ABP): How about AFS workspaces for newcomers? Will they still get workspaces, how is it going to be done? User tickets?
      C (Jan): Yes, ticket to AFS support for keeping track and understanding the use cases
      Q (Frank L, BE-CO): FUSEX is not similar to AFS; worried about compatibility; how about storing/manipulating many small files?
      A (Jan): They are different systems (different tools for ACLs, quotas), but standard operations like compiling should now work without issues (EOS might be slightly slower than AFS)
      
    • 16:20 16:30
      EOS status update 10m
      Speaker: Luca Mascetti (CERN)
      EOSPROJECT status update:
      Q (Joel): When is FUSEx available?
      A (Jan): Please let us know if you need it on EOSLHCB; test can already be carried on EOSHOME and let us know if you find problems; for the time being is being kept off as the Wigner machines would introduce latency which might degrade the users' perspective on the service; it can be enabled if really needed, keeping in mind the drawbacks.
      Q (Giovanni): install EOS-FUSEX on a non-CERN linux distribution (Ubuntu)? AFS works.
      A (Luca): if possible, try to stick with the CERN supported systems
      A (Jan): could be a pain to configure (known issue), but packages are available for Fedora and Ubuntu
      A (Kuba): we focus on CERN CentOS/SLC for the time being
    • 16:30 17:00
      feedback / discussion / next steps 30m
      Free discussion:
      Jan: is there anything painful that we are not aware of? Are there outstanding bugs to be reported?
      Q (Frank Locci): setting up platform with eosproject/eoshome; we do see some issues; can we have a technical meeting?
      C (Jan): sure, that can be done
      Q (? BE-CO): still cannot access EOS from TN; we have also some bugs to report
      A (Jan): please do (ticket/technical meeting if needed)
      A (Luca): for the TN, this is not on our side; it is a matter of allowing a set of machine to be accessible in TN
      C (Joel): have issues with services on an SLC5 box..
      Q (Alessandro Di Girolamo): what about ACRON service, is required.
      A (Jan): ACRON should work against EOS (has a workaround); it will probably be re-written but functionality will stay
      A (Manuel): re-writing in progress, but not quite there yet
      Q (Andreas Pfeiffer): Will EOS have readonly non-Kerberos, non-token functionality (= anon access)?
      A (Luca): We will try to make it possible for client machines inside CERN, only for project areas (not for personal space); matter of configuration, pending decision 
      

       

      Next meeting: in  ~3month time.