Attending: Emanuele, Jens (chair+mins), Steve, Dan, Brian, Sam, JohnH, Matt, Duncan, Gareth, Winnie


0. OBP

   Welcome to Emanuele who has joined Glasgow.  He's previously been at NIKHEF and in
   Bangkok and has done work there for ATLAS and CMS but is now joining the dark side
   (admin).

   Operational issue with IPv6 at RALPP, considering turning it off... suggested was due
   to router unable to handle traffic.  Compare with Glasgow's earlier report which was
   due to router or traffic into site - one of those - which one not known.  Central IT
   "overworked" - a not unfamiliar picture.  Glasgow will get new data centre and
   equipment at some point and will obviously ensure that the new stuff can handle
   IPv6.  There is currently only one person handling network configuration across campus...

1. Supercomputing (the conference) is embracing and expanding into
data.  We had wanted to get a speaker to, er, speak about the data side
but have not been successful yet.

2. Updates on ongoing tasks: IPv6, Rucio, DOME, EOS (non-datalake)
testing, DataLake (CERN Flavour). Docs.

   This XDC thing: haven't had time to look at it but we are thought to follow it
   indirectly at least through DOMA?  Need to look more closely when we have more
   time.

   The IPv6 issue above - 100% of CMS pledge (this year's) on IPv6, 80% of ATLAS.

   DOME (Matt) - works.  We are now not recommending that sites not upgrade.
   Which is obviously not the same as recommending that sites upgrade.  If you want
   to get it in this (calendar) year, now would be a good time... Alessandra upgraded
   a test SE.  Raul is believed to have one, although we haven't heard from him in
   a while.  Upgrades depend on how you manage config; you should be using puppet but
   Matt upgraded and configured by hand - however, YAIM is not recommended.
   All sites should upgrade by June '19 by when non-DOME DPM goes out of support.

   Rucio - is it stable enough in the UK that we can offer it to our users?  How
   should we offer it (if we offer it), we need a tutorial that includes deployment
   (CVMFS or Docker) and basic usage - see link in chat log.

   As regards the EOS DataLake (see link in chat to HEPiX presentation), Brian
   has requested access to dteam so we can try it out.  EOS-DataLake is also
   believed to have been set up as a DOMA TPC endpoint.

3. If there are any experiment-requested changes over the next three
months it'd be good to know about them now

4. If there's time (if not we postpone till next week), round table
update - been a while since we've had one.

   Postponed till next week...

5. AOB





Daniel
i guess that one good thing that our central IT buying cisco Nexus 7Ks
cost a lot but seems to work for us with ipv6
at least 20% of ALL QMUL traffic going via IPv6
DP
Duncan
Is ipv6 enabled on non-grid hardware at qmul?
https://rucio.cern.ch/
https://rucio.readthedocs.io/en/latest/installing_clients.html
DR
Gareth
https://rucio.readthedocs.io/en/latest/cli_examples.html
So there appears to be a demo
https://rucio.readthedocs.io/en/latest/installing_demo.html
GR
Brian
https://indico.cern.ch/event/730908/contributions/3153341/attachments/1731117/2797839/HEPIX2018-WLCGDatalakeProtoype-v3.pdf
BD
Daniel
IPv6 enabled on students dexktops etc... at QMUL
DP
Today at 10:30 AM