DTEAM minutes 2009-05-26 ======================= minutes: DTR There was poor sound throughout the meeting. Jeremy could not connect and as a consequence Brian chaired the meeting. LHC VO reports ============== LHCb - no representative CMS (David) - hit by extended downtime at Tier1. This week, trying pre-staging (James), currently discussing strategy. CMS STEP09 starting 1 June for 2 weeks. Discussion regarding ATLAS's experiences of pre-stage testing. Brian said ATLAS pre-staging tests had gone well at RAL. David: how was this done? locally, with SRM commands? ATLAS (Brian) - Preparing for STEP09. Trying to get space tokens ready for data import. Following space required: DATADISK available, space needed,spare,site 80.85 24 56.85 UKI-LT2-RHUL 23.04 6 17.04 UKI-LT2-UCL-CENTRAL 6.49 24 -17.51 UKI-LT2-QMUL 36.79 34 2.79 UKI-NORTHGRID-LANCS-HEP 19.73 26.5 -6.77 UKI-NORTHGRID-LIV-HEP 16.37 6 10.37 UKI-NORTHGRID-SHEF-HEP 8.05 6 2.05 UKI-NORTHGRID-MAN-HEP 8 6 2 UKI-NORTHGRID-MAN-HEP 56.26 48 8.26 UKI-SCOTGRID-GLASGOW 30 30 0 UKI-SOUTHGRID-RALPPD 12.98 3.6 9.38 UKI-SOUTHGRID-BHAM-HEP 8.38 6 2.38 UKI-SOUTHGRID-CAM-HEP 2.88 21 -18.12 UKI-SOUTHGRID-OX-HEP So QMUL, Liv, RAL-PPD and Oxford are under what is required. Sites please let ATLAS UK know (via ATLAS-UK-Comp list) if they have more space. ATLAS-UK can re-jig space if needed. E.g. QMUL could be taken up by RHUL/UCL-Central etc. On Thurs, ATLAS-UK will check space tokens. [post-meeting update from Brian: the following as of 2009-5-26 are the numbers that ATLAS intend to send to sites during STEP09 IF sites do not change their space tokens in the next 24 hours. These are size to be filled and the proportion of data this represents. 52TB 43% UKI-LT2-RHUL 12TB 10% UKI-LT2-UCL-CENTRAL 6TB 5% UKI-LT2-QMUL 36TB 30% UKI-NORTHGRID-LANCS-HEP 18TB 15% UKI-NORTHGRID-LIV-HEP 12TB 10% UKI-NORTHGRID-SHEF-HEP 6TB 5% UKI-NORTHGRID-MAN1-HEP 6TB 5% UKI-NORTHGRID-MAN2-HEP 48TB 40% UKI-SCOTGRID-GLASGOW 30TB 25% UKI-SOUTHGRID-RALPPD 6TB 5% UKI-SOUTHGRID-BHAM-HEP 6TB 5% UKI-SOUTHGRID-CAM-HEP 2.4TB 2% UKI-SOUTHGRID-OX-HEP ] MC will produce 9TB which will go in same ratios as DATADISK. Merging of AOD also going well. Unmerged AOD will be deleted. HammerCloud tests - aim for atlas is to have 10k jobs running in total. Noticed BHAM and Manchester had clusters not being used. Site performance ================ Link from Jeremy. UCL-Central having problems running jobs. Also Tier-1 - possibly related to Castor downtime. Also PPD and Cambridge had problems last week. Tier-1 move =========== What needs to be done? Going back to last week - getting Imperial and Glasgow WMS to take over from RAL. Alessandra: WMS, LB and myproxy need to be set up in UI. Possibly also BDII needs changing in UI and WN. Are sites prepared for this or not? Regional Operations =================== First line support - what is the definition of first line support. Devloping and on-duty rota - agreed to use a google calendar? Also no clear decision on which tool to use. Helsinki meeting - Pete possibly going. action: Jeremy - to set up google calendar and to progress decision on which chat tool to use. There was a discussion on whether the latest version of skype supported chat. VOs wrt NGS and GridPP ===================== - RTE tags Steven Young, J. Churchill and David Martin from NGS joined the meeting and described the problem. Glue publishing of software versions - what publishing is needed on gridpp/glite sites. Glite - generic info provider doesn't allow multiple plugins to publish into the same glue DN. GIP plugin developed by NGS but can't be dropped in. Options include to submiting a patch to glite or link to a directory. action: Dug to write up the possible solutions. - Large number of small VOs We were not sure what was meant by this. - Shared applications suites (between VOs) NGS has uniform execution environment - fixed location across sites. Alessandra: GridPP uses the VO_NGS_AC_UK_SW_DIR environment variable and this should be used. Difference perhaps is that in the NGS it is the site's responsibility to install software where as in GridPP it is the VO's responsibility. We need to agree on a tag format. Objection to the use of VO_NGS_AC_UK as it may not exist in the longer term. Alessandra: tags are dynamic and can be removed. Tickets, actions and issues review ================================== Only one: 46024 - publishing DNs accounting https://gus.fzk.de/ws/ticket_info.php?ticket=46024 What tool should we use for actions? Site issues =========== Glasgow: Camont jobs use up too much memory. Karl has killed jobs and is investigating. AOB == How to publish my machine architecture - action on Tier-2 coordinators to diseminate this.