Minutes 8/13/18
Steve:
Hitting the voms-admin server—why
Nobody knows—Steve will send around a email with the
IP’s that are hitting.
QMUL= Queen Mary University London
No known DUNE collaborators Daniel Traynor is the admin.
Ken—mcc11
Big samples are done, 3-4 requests still in the queue. May have another round of big ones.
A bit more than halfway. Every UK site except Edinburgh has contributed cycles
Biggest issue—memory settings in the factory entries
Imperial 8GB 4 cores
Oxford 4GB
Others 2000-2500
Is anyone up to increasing memory to 4GB? Or if you support multicore entries?
For reconstruction uses tensor flow which will use multiple cpus, will request at least 2 cores.
Heidi—multiple stage reconstruction—because of way ART works it doesn’t drop memory like it should. G4 part is bigger one but won’t drop memory after it goes to deism.
RAja—at RAL you can request what you need and the queue will give you that
Ken—can definitely use 2 cores, some can do four.
Edinburgh we are getting glideins but they are not reporting back yet, Ken will diagnose.
0——
Robert—can import existing SAM datasets into Rucio,
Have done that with one mcc10 data set.
Rucio has copied that to liverpool
Tried Manchester but we seem to have lost ability to write.
Have been using SRM.. next step is to try with webdav.
Using fts3public server at cern.ch at the moment.
Raja—RAL Echo storage is in downtime right now and will stay in downtime until midday tomorrow.
Next step to add RAL as a Rucio SE
Hoping James could start working on some of this
James—have accounts, not yet able to connect to Rucio server last time I tried.
Still dealing with the rucio web service exemption.
Still have to install the rucio client on your own.
Client only works with python 2.7
Rucio has a lot of external dependencies on rpms