Notes from GridPP EVO meeting 2011-08-17 Attending: Edinburgh: Wahid RHUL: Govind Glasgow: Sam Liverpool: John RAL: Brian, Jens Sheffield: Elena T2K: Jon Lancaster: Matt PMB: Pete Apologies: QMUL: Chris 20475 - assign to Sam - can possibly do more efficiently than T2K script 20474 - assign to Wahid 17931 - soon to arrive DPM will partly alleviate this, needs careful testing 16729 - Chris has a StoRM dumper, close 16728 - What is in the wiki? Edge cases which should fail, not things which should work. 16727 - Plan exists, now needs following up. 16722 - close, done. See also Wahid's monitoring. 16721 - Atlas job failure rate due to SE is improving. Sam will turn on job recovery at Glasgow - write to local storage on WN to retrieve data if writing to SE fails. Can run another (recovery) job to retrieve it. https://savannah.cern.ch/task/?group=srmsupportuk Bristol StoRM happiness? Maybe StoRM is less lightweight than DPM? Bristol "absorbed" by Oxford. The logic of the script should prevent you from dpns -r. Shouldb't be a race condition - directories should not be used. Can the VO do it themselves? There may be an lcg-command to do it. 4,000,000 empty directores at T1. Why did they end up with this many empty directories? They need to do lcg-del on the directory but don't always know if the directory will be empty. Test DPM at Glasgow: can have empty directories which we can try deleting. Can run out of inodes if not doing this. What is the subdirectory limit on DPM filesystems? 999,999 subdirectories in a directory. broke the LFC. Moving things into space tokens: Wahid will try to do this, but will check with J.-P. Will run a general test, not on T2K data. [09:59:48] Wahid Bhimji joined [10:00:10] Wahid Bhimji hi- I can't find headphones so may be mr echo [10:02:07] Wahid Bhimji have we really been talking about this since May - gosh ... [10:02:22] Elena Korolkova joined [10:04:39] Elena Korolkova which problem do you see in Sheffield? [10:04:39] Jonathan Perkin joined [10:05:53] Wahid Bhimji Its done [10:06:04] Wahid Bhimji indeed chris did it [10:07:09] Wahid Bhimji Storm 1.6 upgrade - [10:08:56] Matthew Doidge joined [10:11:02] Wahid Bhimji yeah - I can see if Bob Cregan is maintaining it [10:11:23] Pete Gronbech joined [10:15:14] Wahid Bhimji oh yeah I remember [10:16:23] Elena Korolkova I think wn's disk space at QM will fill very quickly [10:18:55] Wahid Bhimji If you have a link to a talk or something where it is described brian that would be good. Seems a lot of infrastructure for something that wont be that useful (the WN saving file thing) [10:20:00] Wahid Bhimji in this case ATLAS don't care if it deletes files within that directory [10:20:36] Wahid Bhimji but as I said when this happened last time - it is crazy for atlas to have to ask for such things - big risks [10:20:53] Wahid Bhimji is there no SRM commands [10:31:19] Jonathan Perkin left [10:31:20] John Bland left [10:31:22] Elena Korolkova left