Attending: Jens (chair+mins), Govind, John B, John H, Dan, Steve, Pete, Brian, Sam

Apologies: Alastair


0. Operational blog posts - new quarter!  We also need some goals for this quarter. Suggestions?

   Maybe something on CVMFS - is it worth testing any of the new stuff?  Maybe, but need to speak to
   Catalin first.  (New stuff is security)

1. Whatever happened to ATLAS accounting

Alastair writes:

\begin{Alastair}
There is an ongoing discussion within ATLAS about this at the moment.  The wiki page can be found here[1] (although it is
probably ATLAS only even though it has been requested to be made public).  The json format for publishing information has
been agreed for some time.

The US cloud has been switched to using the non-SRM accounting.  The plan is to talk to the UK cloud (us) about SRM
decommissioning next. There is some confusion about how things will work with DPM.  Wei Yang thinks that DPM 1.9 or later
will provide the json via http where as Cedric thinks it will be provided directly via PROPOUND.  I am guess this will go
back to the developers shortly.

As of today, there are two probes running, one has “gsiftp” scheme fixed[2], the other has “srm” scheme fixed[3].
This means, that https shall not work until ATLAS upgrade various scripts (2 - 3 week ETA) regardless of how DPM does things.

[1] https://twiki.cern.ch/twiki/bin/viewauth/AtlasComputing/SRM2GridftpMigration
[2] https://gitlab.cern.ch/rucio01/rucio/blob/next/tools/probes/common/check_gridftp_space#L59
[3] https://gitlab.cern.ch/rucio01/rucio/blob/next/tools/probes/common/check_srm_space#L53

P.S. Sorry I can’t join the meeting at the moment, otherwise I could have said this in person.
\end{Alastair}

There is a need to communicate with the experiments - which is why we normally have experiments' representatives,
and if they don't have time to talk to us then that's OK if they're happy but not OK if they're not.

Like the WebDAV deletion which causes the accounting to become inaccurate - did we only discover this post hoc
(or indeed propter hoc, after Lisbon), or was it a case of miscommunication?

Also a case for using the wiki here.  Partly that the documentation for us should be in the wiki but conversely
also that the wiki should be up to date(!) so the VO (and others) know what to expect.

https://www.gridpp.ac.uk/wiki/GridPP_approved_VOs
https://www.gridpp.ac.uk/wiki/GridPP_VO_Incubator

2. GDB

https://indico.cern.ch/event/578985/

--- the main message for us is that experiments seem to be increasingly relying on containers, so, will need to access
storage from whatever is in the container, and they will be relying on CVMFS to distribute the images.  (It's like back
in the very early days of the grid where LHCb were copying tar files around with libraries etc., only with more isolation)

and preGDB (note the LIGO and data issues talks)

https://indico.cern.ch/event/578969/

On the subject of LIGO,
\begin{Alastair}
Since the end of last year, I have been keeping in touch with Brian Bockelman about his work getting small VOs to access their
data through CVMFS.  This is what LIGO have used heavily.  The thought is that now might be a good time to try this at RAL,
we can use S3 as the backend for the data and Catalin has done most/all the necessary upgrade work.  I have arranged a meeting
with Brian in HEPiX next week to discuss this.  If anyone else is there and would like to also join they are more than welcome
(I’ll add to email thread).  I was hoping that this was something I might report back to the storage group in about a month.
\end{Alastair}

One other interesting thing about LIGO is that they are interested in secure CVMFS, so maybe scope for testing this; note
that security would normally disable caching as you would not be able to inspect and cache at the HTTP level.  Also they
are interested in authentication stuff, including federated authentication.

Data issues seem to be suggesting that fewer than half of the SEs are providing 90% of the storage which in turn suggests
a move to fewer sites - presumably this would be part of the wlcg-data whatsit group/task force in which GridPP is already
represented.


3. KeyDocs - yes we still have nasty red ones for storage.  And remember others are allowed to edit them too.

   Reminder - most of the very red ones are storage ones!

4. AOB

   What happened to the IPv6 stuff?  Sometimes test SEs mess up performance numbers and some of the IPv6 test stuff
   might do it, too.  Sam mentioned Glasgow turned off their IPv6 SE as it was not seeing good performance due to
   IPv6 not being supported end-to-end, i.e. there was some hardware somewhere which was unable to route IPv6 efficiently.
   See Dan's link in chat.


Daniel Peter Traynor: (19/04/2017 10:00:59)
yes
Jens Jensen: (10:01 AM)
CAn you hear me?
Daniel Peter Traynor: (10:28 AM)
some lustre news: http://www.theregister.co.uk/2017/04/18/intel_loses_its_lustre_bins_ownbrand_hpc_filesystem/