Choose timezone
Your profile timezone:
PASTE meeting
→
Europe/Zurich
Description
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, 17th July 2007, CERN
Chair: A.Tsaregorodtsev
Minutes: S. Paterson
CERN: M.Bargiotti, G.Castellani, C.Cioffi, E.Lanciotti, S.Paterson, R.Santinelli, A.Smith
Phone: R.Bernet, N.Brook, R.Graciani, L.Magnoni, R.Nandakumar, M.Sapunov, A.Tsaregorodtsev, R.Zappi
Apologies: P.Charpentier, J.Closier
1. Production Status Report
Minutes from the production meeting are available here.
Nick gave a summary of the production status, all of the A1 physics group requests have been completed. A potential
problem was identified with the workflows corresponding to production IDs 1879, 1880. Cedric agreed to investigate
whether this was really a workflow problem or site specific issues. ( ACTION: Cedric )
The staging progress at GridKa has been slow with many jobs failing to stage all data in a 24hr period. Johan agreed
to monitor this issue. ( ACTION: Johan )
The recent SAM problems reported by Roberto and Stuart have resulted in a slowly decreasing number of available
sites. Stuart and Roberto clarified that after sitting together last Friday, problems were observed with the publishing
of SAM results. Bugs have been submitted to Savannah regarding this. A proposed solution is to run SAM jobs via
DIRAC and publish our results back to the central SAM database in the same way. It was agreed to pursue this
solution and arrange some discussion with the SAM service providers (for the Thursday Operations meeting). (
ACTION: Stuart, Roberto )
Nick commented that the stripping appears to be failing in DaVinci. While the stripping is supposed to be running with
v19r1 it appears to get the environment setup for v19r3. This is to be investigated and understood. ( ACTION:
Stuart, Nick )
Raja asked whether there was a problem on two of the VO-boxes where the proxy renewal mechanism did not appear
to work. Ricardo conveyed that Joel's proxy had expired and Raja was unsure whether his credential was uploaded to
the MyProxyServer so this should hopefully resolve the issue.
2. Production Operations Issues
Andrei reported that the fix for the sandbox problem at IN2P3 was in place and tested at CERN. This is in v2r15p2
which will be deployed on the Production and User systems later today. ( ACTION: Andrei )
Stuart mentioned that the activity of updating the Stager functionality to SRM2.2 was pending the release of the
python binding to GFAL. Andrew commented that after communicating with the developers, one release is in
certification and expected at the end of the month. Nick suggested that we should expedite the delivery of these
tools as they can affect our testing schedule. Andrew agreed to sit with the developers. ( ACTION: Andrew )
Andrew remarked that the storage endpoints for the SRM2.2 tests at the Tier-1's are pending. Hopefully some
progress will be made by next week. ( ACTION: Andrew )
Ricardo mentioned that emails had been sent regarding the use of edg-brokerinfo in the pilot script. Gianluca agreed
to resolve this for v2r15p2. ( ACTION: Gianluca )
3. StoRM Tests at CNAF
E.Lanciotti presented Functionality and Stress tests on a StoRM Instance at CNAF.
There was some discussion regarding the StoRM functionality and Elisa clarified that any file system (not only GPFS)
can be used. Andrei suggested calculating the CPU / Wallclock ratio for comparison to other solutions under load.
More details on the GPFS implementation at CNAF were also requested. Roberto conveyed that tests comparing
StoRM to other solutions were performed by Vincenzo and Angelo and the link will be circulated to the developers list.
( ACTION: Roberto )
4. Moab-based Priority Calculation
Gianluca attached a presentation about Moab to the agenda page.
As opposed to Maui, Moab has a service interface which may be utilized. Gianluca confirmed that the memory depth
for jobs in the system is configurable. Job.txt file is updated continuously with the job status and other information,
Moab then takes this file as input and calculates a priority for each job. Tests on the scalability and speed of Moab
when tens of thousands of jobs are in the task queues should be performed. Ricardo suggested using ~10K jobs and
frequent reloading of the Jobs.txt file every minute to see how Moab reacts. Ideally a curve up to 100K waiting (plus
10K running) jobs is desirable. Moab is a commercial product so the licensing issue should also be cleared up. (
ACTION: Gianluca )
5. AOB
Andrei mentioned that his vacations will start next week and suggested Stuart run the PASTE meetings during his
absence. Since we are entering the middle of the vacations period some PASTE meetings could be missed although
this can be decided after next week's meeting. Ricardo suggested that since the SRM2.2 tests will be ongoing, we will
most likely require the meetings to coordinate this activity.
Carmine reported that the Bookkeeping receiver was getting many duplicate XML files about already existing replicas.
This is a problem for the Bookkeeeping since the same files must be processed several times causing slowdown.
Andrei requested some example requests from Carmine and agreeed to investigate this issue. ( ACTION: Carmine,
Andrei )
There are minutes attached to this event.
Show them.