ADC Weekly

Tuesday 18 Jun 2013, 15:40 → 17:30 Europe/Zurich

3162/2-E01 (CERN)

I Ueda (Department of Particle Physics-University of Tokyo)

15:40 → 16:15 Hot topics

15:40 SW Installation system status

The releases tagging is now done by the new one with panda jobs since June 11.

mail to atlas-adc-operations https://groups.cern.ch/group/atlas-project-adc-operations/Lists/Archive/Flat.aspx?RootFolder=%2Fgroup%2Fatlas-project-adc-operations%2FLists%2FArchive%2FSwitch%20of%20ATLAS%20SW%20Releases%20Validation%20System

15:45 reproducing lost files

Speakers: David Cameron (University of Oslo (NO)), Sasha Vanyashin (Argonne National Laboratory (US))

Slides ADCWeekly180613_lostfiles.pdf

15:55 T1_datadisk

Speaker: Tomas Kouba (Acad. of Sciences of the Czech Rep. (CZ))

Slides 20130618_Kouba_T1DATADISK_status.pdf

16:05 Storage Area Automatic Blacklisting (SAAB)

Speaker: Dr Salvatore Tupputi (Universita e INFN (IT))

Slides https://twiki.cern.ch/twiki/bin/viewauth/Atlas/StorageAreaAutomaticBlacklisting

SAAB

• to be activated tomorrow with an elog and a link to the document
• SSB major views to be updated to include the SAAB column
• 'SAAB actions' log file to be prepared (similar to panda 'incidents' page)
  • the per-site history view does not fulfill the request

16:15 → 16:25 rucio naming convention for Panda jobs

Speaker: Stephane Jezequel (Centre National de la Recherche Scientifique (FR))

Slides ADC_rucio.pdf

16:25 → 16:40 AMOD report

16:40 → 16:50 Shifters procedures on issues related to FAX

Speaker: Robert William Gardner Jr (University of Chicago (US))

more information https://indico.cern.ch/conferenceDisplay.py?confId=258240

• A.DiGirolamo, R.Gardner: discussions on the procedures for the shifters on-going
• R.Gardner: activation of 'allowfax' for the US and some UK sites should have negligible impact. only small scale tests is planned that should not pull attention of shifters. no plan such as to set data server down and trigger a large scale failover to FAX.
• What would happen if a site storage goes down (not by intention as mentioned)? Then, with 'allowfax', jobs after the first failures (usually 2, set on schedconfig) will start accessing files via FAX. FAX will try to get another copy of that file first within the cloud if they are available, and if not, outside the cloud, again if available, only on sites which participate to FAX.
  If the storage is not completely available (i.e. not even for writing), then the jobs will fail while trying to upload the output, but if the storage had just a glitch, then the jobs will be successful.
   

16:50 → 16:55 AOB