PASTE meeting

Tuesday 5 Jun 2007, 10:00 → 12:00 Europe/Zurich

Andrei Tsaregorodtsev

CERN, bat 2 R-030 To join the telephone conference, click here: https://audioconf.cern.ch/call/0122882 To dial in to the conference: a. Dial +41 22 7676000 b. Enter access code 0122882

Production and PASTE Meeting, 5th June 2007, CERN Chair: A.Tsaregorodtsev Minutes: S. Paterson CERN: M.Bargiotti, G.Castellani, P.Charpentier, C.Cioffi, J.Closier, A.Maier, U.Marconi, S.Paterson, R.Santinelli, A.Smith, A.Tsaregorodtsev Phone: R.Bernet, A.Casajus, R.Graciani, G.Kuznetszov, R.Nandakumar 1. Production Status Report Minutes from the production meeting are available here. Joel mentioned that simulation has restarted, there was a software installation problem where tags were not published but this has since been resolved. Andrei has now released the 'on hold' jobs after enabling the job software tag requirements. The Transfer Agents were stalled but are now operating normally. The issue of banned SEs (in this case not having been updated for several months) was discussed. It was agreed to put Marianne in charge of 'unbanning' the sites as necessary and a simple Twiki page should be set up to facilitate logging of these actions. ( ACTION: Marianne ) Joel has the next production request (several thousand events) from Olivier which will require a special generator and new workflows will be created for this. ( ACTION: Joel ) The Reconstruction activity is ongoing, at CERN there are still problems due to lost tape requests blocking the system. There will shortly be a Castor tape intervention to upgrade the stager. At PIC there are very few running jobs, Ricardo made us aware of a public holiday in Barcelona on Monday and expected further investigation today. The final version of DaVinci for the Stripping will be v19r1. Joel's test jobs have passed through the Stager and are waiting. There is one for each of the five sites for which the stripping is eligible (none for NIKHEF / IN2P3 since no data available). Joel reported that his certificate had become mapped to SGM, without warning, at several sites. This caused two weeks of development time to be lost. GGUS tickets were submitted but this is a YAIM related issue. Stuart suggested having SAM tests to check site user mappings and Roberto agreed to investigate this. ( ACTION: Roberto ) Philippe agreed to send a summary to this afternoon's Management Board regarding the following issues that are delaying the LHCb production activity: - SAM system breakdown - Change of user mappings - CERN CA not being recognised at dCache sites ( ACTION: Philippe ). 2. Production Operations Issues The LHCb LCG user access policy was discussed after a summary of the recent user behaviour observed on the Grid. Andrei suggested making a strict 'charter' when joining the VO such that all LHCb jobs will go through DIRAC. Stuart mentioned that this is the only way by which we can guarantee accurate VO accounting information. Philippe agreed to make up a draft policy for LHCb Grid usage and stressed the point that all possible use-cases should be actively supported. In order to control the 'backdoors' it was suggested to either restrict the submission of jobs by host (e.g. WMS machines) or more simply by DN in the the generic pilot regime. The policy for this should be made clear to users. There was some thought about how best to communicate this to the collaboration and it was agreed for Joel to add this to his LHCb week presentation. ( ACTION: Joel ) For the LHCb packaging of Root it has been decided to proceed with Ganga resolving requested Root versions into corresponding DaVinci versions. Ganga can then submit a self-consistent script via the DIRAC API to be executed in the DaVinci environment for use on the Grid. This makes use of a script from M.Clemencic allowing Ganga to determine the DaVinci - Root mappings. In this way we prevent any downloads of Root via Ganga to the WN. Andrew M. reported that job splitting taking sites into account is on the way and should be ready this week. Stuart raised the issue of recent user output sandboxes in Ganga and questioned whether the automatic download of job outputs could be turned off via a flag. Andrei made the point that users should be aware of the possibility of using the DIRAC output data mechanism. Ganga will use the DIRAC API for SE interaction and users can therefore specify their preferred destination site via DIRAC SE name. For the DC06 cleanup Marianne has started the DIGI removal and reports slow progress with the bulk removal tools. Roberto reported that a reminder was sent to the remaining Tier-1 contacts for the physical removal of files. Marianne is preparing the lists for DC04 CERN_Castor SE name removal in the catalogue and mentioned that CERN_Castor was successfully removed from all DC06 files. Stuart asked about the other problematic names e.g. PIC_Castor, CNAF_Castor and Marianne confirmed these are also to be removed. Fixing FC fields with malformed entries e.g. carriage returns, is also in the pipeline. ( ACTION: Marianne ) Andrew highlighted that transfer rates were peaking at 100MB/s and directed us to this link for monitoring the situation. Raja mentioned that SLC4 will be deployed at RAL in July. This should be fine for DIRAC assuming 32bit mode and correct resolution of the CMTCONFIG variable. CASTOR at RAL is currently down and it was agreed to pursue migration with low priority until the system is more stable. The possibility to integrate SAM tests / detailed Pilot failures and RB throughput reports in the Dashboard was discussed. This would be advantageous for reporting the stability of essential Grid services and could utilize other DIRAC components for a fuller picture. Stuart reported that further to Philippe's request from the Monday Production meeting, root protocol is now in place in the Job Wrapper. This will be part of the v2r15 release but is currently pending until test jobs have completed. 3. DIRAC3 Tasks Andrei commented on the slowdown of clients when contacting the CS. Ricardo mentioned this was an initialization problem and Gennady asked for some examples of how to do this. Adri reported that a patch had been committed to CVS. Stuart conveyed that the discussion of the Job Wrapper had been fruitful and development is slowly starting. Andrew's DMS streamlining is ongoing. 4. AOB Carmine has until the end of August and it was agreed that continuity should be assured with the Bookkeeping. Philippe reported that some high-level discussion had occurred about how to access DPM / Castor files from inside a Root library. Further progress on this issue will be reported next week. ( ACTION: Philippe ) Adri raised the issue of SLAs for the VO-boxes at the Tier-1 sites and suggested some discussion of this at the next PASTE meeting. It was agreed that Adri should look through the existing document and check if expected service requirements are reasonable. To help this, Joel agreed to send the CERN requirements to Adri. ( ACTION: Adri, Joel )

10:00 → 10:15 Production status report

- Report from the Production meeting:

10:15 → 10:45 Production operations issues

- User LCG access policy - DC'06 cleanup update - LHCb ROOT packaging to be used by GANGA - Job splitting with account for sites - Dashboard strategy

11:05 → 11:35 DIRAC3 tasks

Round table report

11:35 → 11:45 AOB