Attending: Jens (chair+mins), Steve, JohnH, Sam, Dan, Govind, Teng, Brian, Winnie, Matt, David


0. Operational blog posts

(still no new blog posts!)

Daniela's mail didn't seem to attract any responses but the faulty permissions is either due to a xroot-only problem of some sorts,
or in DPM (where directories are normally created through SRM) the default permissions can be configured.

1. hepsysman has been announced; it's in everybody's favourite
Glaswegian location, Glasgow. Did we propose a storage hands on activity
in or along hepsysman or was that just an idea?

ZFS and xcache are probably too specific to handle in a workshop.  Maybe there is some value in getting people together in the same room
to hack some stuff, too.
- storage accounting?
- IPv6? but people probably already know what they need to know, and didn't we do it last time?
- storage service certificates (like multi-SAN)
- data transfer zone revisited?
- xrootd configuration.  Teng could volunteer (in the original sense of the word) to give a presentation.

hepsysman is 16 Jan.

2. Which reminds me about the cloud workshop at Crick. GridPP was well
represented last time, and we attempted to do some Real Work(tm), to
move data between GridPP and a public cloud. Deadline for this year fast
approaching. Discuss.

Still needs some ideas; deadline is the 8th so plenty of time... and the workshop itself is 8 Jan 18.

3. Round up of loose ends - DPM 1.9 upgrades?  CERN@School? SciTokens? 
xcache updates?  storage accounting?

Bristol running DPM1.9; also Matt reports much faster accounting, so that means the problem is fixed, hooray.

CERN@School - to Jens's surprise(?) they're submitting to a CE in the good old fashioned grid style, running at QMUL.  Used to be
able to match against CEs advertising GPUs, as they use GPUs to optimise their TensorFlow [interestingly it's the learning that is
optimised not the application phase because Google's hardware thingamajig did the opposite.]  They had lots of small files which can
conveniently be untarred.  So this sounds a bit like a(nother) success story for GridPP - maybe we should make a bit of noise about it
(or at least a blog post?)

Also note we have a gridpp-support mailing list for solving problems and summarising/describing solved problems.

4. AOB

Bug in DPM's SRM reported.



Daniel Peter Traynor: (06/12/2017 10:17:47)
https://www.eventbrite.co.uk/e/research-councils-uk-cloud-workshop-tickets-39439492584
ithis meeting?
jens: (10:18 AM)
yep
Matt Doidge: (10:21 AM)
Our accounting takes about 12 minutes to run, we're on 1.9.0 (legacy)
David Crooks: (10:24 AM)
Sorry I'm late, I got pulled into a meeting
Daniel Peter Traynor: (10:53 AM)
upgrade day for every body next week then? 
Steve Jones: (10:53 AM)
https://www.gridpp.ac.uk/wiki/Problems_After_CA_1.88-1