Tech Meeting Minutes 20110310
From ScotGrid
| Table of contents |
Hot topics
Edinburgh
Andy: Not too much: previously had problem with Cream CE taking up excessive load, which turned out to be MySQL corruption. Andy removed database and created a new one, which has been running fine ever since. Still get some oddities with CPU usage, but suspect this is due to Cream CE middleware. Wahid had been running Hammercloud tests since last meeting, ongoing. Other work; middleware provisioning by ECDF system team, 2nd Cream CE, ongoing. After 2 Cream CEs, will turn off LCG-CE. Question about squid proxies - is there a consensus about frontier/vanilla squid? Sam: Using an older frontier squid, haven't upgraded because of confusion regarding having to build new version from source, etc. We decided not to upgrade until its clear how best to upgrade. People at Atlas looking into this. Oxford were looking at standard squid. Need to follow up on what the UK is doing. Andy: Looking into testing squids
Re: GGUS, 68095, WLCG - Andy not sure what they were looking for.
Sam: DPM Publisher has a legacy/non-legacy mode. Try looking in information system on DPM. Is it calling dpm-listspaces with --legacy flag? Try removing --legacy flag. In /opt/glite/etc/gip/provider, se-dpm.
Glasgow
David: Not too much: issue with Cream CE svr014 which Stuart was looking at, issue temporarily resolved [Sam: Stuart looking into it, database issues] Of note, be careful about marking LCG-CE into downtime once it's draining. May be better to remove from GOC DB at start of drain.
Durham
Peter: Finally received new worker nodes, x5650s, installation going OK. Going through tickets, version of YAIM causing problems. Cream CE on tasklist. Looking at Cream CE, going through tickets which are taking time. 68094 (Mandatory WLCG InstalledOnlineCapacity not published) closed. Previous ticket on number of logical CPUs closed as well; previous calculation on number of logical CPUs was wrong. On YAIM, have raised ticket. Sam: What version of DPM are you using? Peter: During solving another ticket, removing spacetoken for ATLAS, upgraded gLite 3.1 on DPM. Sam: with DPM, conflict between difference files shouldn't have happened with DPM not of version 1.6. Peter: Problem, groups.conf should be converted into mkgridmap, not happeningcorrectly. Sam: SE vs CEs; SEs don't worry about user level mapping but group level mapping. Peter: Re: Bugs in YAIM, posted DMSU ticket in chat window Sam: Wahid tested for last version of DPM on gLite, 1.8, which had a memory leak which was a 3.1 specific issue. Pragmatic solution was to upgrade to 3.2/SL5. Peter: May take mkgridmap file from CE and transfer it manually. Sam: In general, for storage issues feel free to contact GridPP Storage. Peter: MCDISK ticket done [waiting for Brian, then closed]
AOB
Andy: Security Challenge, any updates? David: No updates, still keeping a watching brief. Andy, David and Sam: Dashboard project; Edinburgh have a new student coming in in the summer - need to keep in touch to dovetail dashboard projects and look where things are complementary. Jamie Ferguson working on revised dashboard which can be made more generic.
Chat transcript
[11:02:08] David Crooks http://indico.cern.ch/conferenceDisplay.py?confId=130812
[11:11:20] Peter Grandi There are several details on the DPM and other BDII issues in the Durham tickets.
[11:12:09] David Crooks Thanks - we'll come back to you after ECDF if your audio is working better
[11:12:39] Peter Grandi in our case there were TWO provider scripts for DPM, one is 'se--dpm' as SamS weas saying
[11:14:08] Sam Skipsey /opt/glite/etc/gip/provider/ is where all the DPM scripts are.
[11:14:25] Andrew Washbrook righto - thanks. I will look into it
[11:14:36] Sam Skipsey se-dpm is the one which calls dpm-listspaces (and is the bit that generates GlueSACapability lines)
[11:16:33] Andrew Washbrook @Sam: in se-dpm there is /opt/lcg/bin/dpm-listspaces --gip --protocols --basedir home --site UKI-SCOTGRID-ECDF
[11:16:56] Peter Grandi https://gus.fzk.de/ws/ticket_info.php?ticket=67884
[11:17:48] Peter Grandi https://gus.fzk.de/ws/ticket_info.php?ticket=68094
[11:18:44] Peter Grandi https://gus.fzk.de/ws/ticket_info.php?ticket=67440
[11:23:02] Peter Grandi https://gus.fzk.de/ws/ticket_info.php?ticket=68470
[11:23:56] Peter Grandi https://gus.fzk.de/dmsu/dmsu_ticket.php?ticket=68392
[11:28:42] Sam Skipsey Hi Andy - that looks good, so it's not the basic configuration that's at fault. Try running the query itself and checking what it produces.
[11:28:51] Andrew Washbrook ok will do.
[11:30:44] Peter Grandi BTW very useful survey of host adapters at RAL: http://www.gridpp.rl.ac.uk/blog/2011/01/12/sata-raid-controller-experiences-at-the-tier1/
[11:35:28] Peter Grandi thanks a lot
