Tech Meeting Minutes 20101216
Present: Graeme, Sam, David, Peter, Stuart, Andy, Michael Apologies: Mark, Wahid
- Suffered failure of main switch into IPPP - networking lost, SRM down ticket from ATLAS. Now recovered, but may need downtime for switch replacement. Also nfs area problems.
- Networking Document
- Information is correct.
- Glasgow: testing to improve utilisation of 6Gb/s shared (at the moment can't really get more than 140MB/s, which is a bit disappointing).
- Edinburgh: 1Gb ok for now; good that network layout is very well understood.
- Durham: 1Gb shared fine for foreseeable future.
- ACTION: Graeme to send this info on to Pete Clarke.
- Holiday Cover
- Best efforts as usual. Use email except if there is a real emergency.
- Edinburgh CREAM CE works - Andy applied a patch used by IN2P3 which uuencodes inputs and outputs. Need to customise the submission script to request 6GB VMEM. New ACTION.
- Graeme noted that ATLAS try to set a shell VMEM limit much lower than this to trigger an athena out of memory error, avoiding a batch system kill.
- Glasgow disk server maintenance: Sam is progressively draining SL4 servers; final decision on ext4 vs. xfs not yet taken. Saw high load on disk045, which is ext4 - ext4 performs better in later kernels.
- Edinburgh analysis priority - no progress.
- Edinburgh squid/cvmfs - to be discussed with Orlando; also use f2f meeting.
- Glasgow pool account cleanup scripts - not yet done.
- Edinburgh glite-APEL: have a VM, but not yet installed.
- Durham glite-APEL: have a VM, but not yet installed.
- There may be a 1 month stay of execution for R-GMA based accounting.
- Glasgow analysis jobs now prevented from running on old CV kit.
- Glasgow GFAL_LCG_INFOSYS now fixed.
- Durham biomed VOMS fixed.
- shift.conf issue was raised with DPM devs.
- Next meeting Jan 6: http://indico.cern.ch/conferenceDisplay.py?confId=117463