Training sysadmins Course on 2013-11-25 and 2013-11-26.

• Date and time: 2013-11-25 10:00-16:30 / 2013-11-26 9:00-12:00
• Place: CSCS / Meeting room first floor.
• External link / EVO:

Agenda

• item
• AOB

Attendants

• CSCS: George Brown, Miguel Gila, Gianni Ricciardi
• PSI: Fabio Martinelli
• ETH: Daniel Meister
• UNIBE: Gianfranco Sciacca
• UZH: Roland Bernet

Minutes

• item

Action items

• Unmount experiment software on WN79 as test - modify node health check
• Email ticket details for APEL parser
• Update ATLAS webrt tickets - Webdav/ cleanup
• CMS old files list in dCache
  • ( Fabio ) : when CSCS will insert in their Chimera DB the view v_pnfs they will be able to quickly generate reports like: https://wiki.chipp.ch/twiki/bin/view/CmsTier3/NodeTypedCacheDB#Space_usage_by_Year_of_Last_Acce
  • ( Fabio ) : I tried the new PG 9.3 materialized view feature to materialize the view v_pnfs, it worked, once materialized it get faster to run the queries while without the materialization each query is going to rebuild from scratch the view v_pnfs.
• Multicore jobs - ARC
• Check job by job description rather than job ID
• PSI UI for external checks
  • please list the desired account names and shells : ... , ...
    • gbrown; zsh if installed, if not bash is fine
• CRL nagios check
  • PSI check

  • # cat /etc/nagios/nrpe.cfg.d/check_file_age_cern_crl.cfg 
    command[check_file_age_cern_crl]=/usr/lib64/nagios/plugins/check_file_age -w 100000 -c 120000 -f /etc/grid-security/certificates/d254cc30.r0
    

• Get Roland back on the UI
  • Ensure key is everywhere (cfengine)
• Update chk commands for dcache (dcache26 rather than 22)
• Update chk commands for cream (chk_CREAMCE_slurm)
  • created chk_CREAMCE_slurm
• Update values for HEPSPEC
• Create alias to easily get local user from sacct
  • Changed default format to jobid,jobname%20,Comment,NodeList,NCPUS,start,end,State%20,User,ExitCode
• Modify node health to check for swap
• Modify node health to update node status, e.g. if a node is drained and then passes ok the reason shoudl refelect when the node was marked offline