DUNE SITES meeting 12/2/2019
Present: McNab, Illingworth, Timm, Brew, Kemp, Herner, Kirby, Schellman, Bauer
Since we shifted FE to do singularity invocation
If you specify OS=SL6 but don’t explicitly specify singularity you want to use,
Wind up with the wrong singularity somehow
At Omaha get into something to our wrong singularity container.
Only problem at nebraska is that sites can force you into singularity of their own choosing
Trying to get users to run analysis grids.
Everybody should be porting to SL7
———————
Ernesto—
Have high HPC at his institute
(Campinas)
Just talked to his people about this.
Easy to install CVMFS
We owe him a wiki page
Within 1 week, they should be able to do test
A few hundred nodes, up to 8-16 nodes
Steve to follow up with Ernesto
——
Andrew McNab—pilot proposal
There is money in UK trying to get hold of:
Exascale data management
Provisioning in general sense.
Experiment-neutral sort of way
Generic pilot factory
Able to submit pilots for any experiment
Same stuff that is inside the dune gpvms.
Have circulated to A. Norman, M. Kirby, H. Schellman
———
NERSC
Ready to put the pedal to the metal.
Several jobs ran successfully
Have to do something about the JOB_EXPECTED_MAX_LIFETIME, SCHEDD puts on hold due to the fact that it isn’t scaled right.
Do we have any special edge services at NERSC—no