dTeam Meeting 2006-10-10 ------------------------ Present: Jamie, Graeme, Jeremy, Jens, Greig, Pete, Olivier, Alessandra, Phillipa, Dave Transfer Tests -------------- Summary page clearly shows tests not yet run. Glasgow still to test - awaiting new hardware deployment. Lancaster need to retest when prod less. Manchester testing right now. Sheffield will retest with new sysadmin getting involved. Need to understand poor write rate. B'ham doing write test now. Bristol, Cambridge and Oxford will test this week. Brunel and RHUL tests should go ahead over the weekend. ULC-HEP,CENTRAL next week? QMUL has tested. Need to put figures in. ScotGrid: 3 tests outstanding. NorthGrid: 4 " ". SouthGrid: 6 " ". London T2: 7 " ". New staff took over from Alice at UCL, holding back things. Staff are overloaded. IC-HEP cannot test now - CMS using dCache extensively. Brunel and RHUL still on 100Mb links. Brunel should upgrade in November? QM is unstable because poolfs dies - real problem because QM's contribution to GridPP is large. ACTION: Need to review notes in wiki on site status, because some may be out of data now. T2Cs. JC: Can these tests be automated? JJ: Dave Kant wrote a script for the storage group to submit jobs to do transfers. OA: Need to know what background traffic is? Gangila could do this. GC: A lot of effort to do this. OA: Different view of efficiency - based on what VOs are doing. Could detect storage bottlenecks from VO use. JJ: You can measure bandwidth for transfers using gridftp - this can be published automatically. JC: We need to load the system to improve it - so active tests not just passive ones. GS: Not enough background traffic from VOs to prove sites at the moment, except IC-HEP. (JJ: need active testing then.) Would need to use SAM infrastrucutre with a Dave Kant like test. Would meet many of the problems found by SAM/SFT tests, so... a lot of work. OA: For LT2 need to most of all to improve the bad rates. ACTION: Summary of issues involved in automating tests - for agenda at F2F on 2006-11-03. SA1 Top 5 Issues ---------------- https://twiki.cern.ch/twiki/bin/view/EGEE/SA1 TCG should use this as an input list for scheduling effort. UKI ROC has voted, but can change still. GS: Would raise 5 (poor error messages) as a real problem issue. AF: Should ask TB-SUPPORT for input. JC: Important to give reasons why issues are important - give context. ACTION: PS to send list to TB-SUPPORT. ACTION: JC to resurrect issues list to add new issues as they arise. Quarterly Reports ----------------- LT2 done. ScotGrid, NorthGrid have all the pieces - just need to fill in the numbers. SouthGrid writing report today. HEPSYSMAN --------- Half a day of talks now. Not so many offers of talks - especially on security theme. DC suggested that it's worth chasing Kostas. Alessandra could present Andrew McNabs AHM talk. JF will not attend. Could schedule a discussion of transfer tests. Talk from Cambridge on condor. NGS talk on security? Steve Corbin? TPM --- Team 7 scheduled for week of 30 Oct - clashes with GridPP. Need to change this. ACTION: Phillipa to raise this. Storage Class Implementation Meeting ------------------------------------ GC attended. Ways to implement different storage classes (DiskXTapeY) at T1 and T2s. Did not discuss site implentation. Still much confusion abounds - even over nomenclature. Intention to have a more concrete discussion before the next GDB. Not so complex for T2s - Disk1Tape0 is only class required. ACTION: JJ is implementing policy for RAL. Will document on the wiki. EGEE storage metrics group has not engaged developers - little progress. DC interested in job submission issues. AOB --- OSG interested in adapting RTM for OSG. AF tested certificate renewal in VOMS - was able to register her new certificate using her old one. ACTION: AF to add to wiki and then to circulate this information to UKHEPGRID list. Should have instructions for users and VO managers.