Attending: Sam (even more remotely), Ste, Dan, Jens (chair+mins), Teng, Winnie, Vip

Apols: There's a Matt-shaped hole in GridPP because he's away on hols


0. OBP

Sam reports an operational problem at Glasgow, where jobs fail to access files due to,
apparently, a race condition which we thought we'd seen before, FTS is trying to access
(place) the same file twice and one thread gets unhappy that it fails and deletes the
file thinking the transfer had failed.  The fault should have been fixed with FTS, so
we should check with RAL whether the fix has been deployed?

Or could it be a higher level service like Rucio which inadvertently schedules two
transfers for the same file, thus causing FTS to genuinely schedule two transfers, and
they just occasionally hit the race condition?

Vip reports a problem in Oxford where ATLAS jobs fail to congregate.  Sam believes it
is a resource contention problem where DPM is busy checksumming files, leaving other
files in the queue running out of resources; a problem reported by Petr from Prague
and fixed in 1.13.

Who in the UK is running 1.13?  We're not sure (without either Matt or the monitoring
we had previously); Prague would be running it.  Vip considers upgrading; maybe Nov.
Sam thinks it should be relatively straightforward, with no major changes or config
changes.

1. For today's call we should hear from someone who attended the Rucio Face yesterday?

https://indico.cern.ch/event/844100/contributions/

Teng was present and gave a summary.  Annoyingly there were no slides uploaded - Teng
says it was mostly a roundtable thing - and we're expecting some minutes to appear.
Teng presented monitoring - same as for GridPP - and will be assisting RAL to set it
up.  There's also a Rucio developers' call tomorrow.  Rohini presented SKA usage, incl
functionality requirements and future plans.  There was a multi-VO presentation with
the original work by Andrew Lister now taken over by Ian Johnson, but there are daemons
still not multi-VO.

Apart from the technical stuff, it'd be interesting if there's a (poss. informal) agreemtn
on how the UK will contribute to the Rucio development, probably depending also on the
uptake of Rucio by IRIS communities - but if we're putting considerable effort into
multi-VOing Rucio (and contributing this upstream), then we're putting at least some
of our eggs into the Rucio basket.

Now let's wait for Alastair to upload the mins...

There will be a Rucio "code camp" hosted by CERN in October.

2. On a not unlreated note, did anyone attend the CEPH day hosted by CERN yesterday?

https://indico.cern.ch/event/765214/timetable/#20190917

Noted it had Stig there (from the IRIS collab.) and Tom Byrne from RAL - and at least
it had slides uploaded... Sam thinks there will be a "technical meeting" on CEPH scheduled;
and we can/should otherwise get Tom to join a Wednesday call.

3. AOB 

  Other than the operational issues, no other business.



Link to Rucio F2F at Coseners: https://indico.cern.ch/event/844100/contributions/

Link to CEPH day at CERN: https://indico.cern.ch/event/765214/timetable/#20190917