Attending: Jens, Teng, JohnH, Sam, Rob, Dan, Steve, Matt, Winnie, Elena, Chris, Pete, Govind, Duncan

Apols: Brian's on leave


0. Operational blog posts

... no new blog posts

Steve asks about the CA issue discussed on tb-support, because it also affects some storage systems because it seems to be a problem
with Argus running on SL6 (and Argus is written in Java, the problem is specific to Java.)  However, Chris had had problems on his
CentOS7 systems only till he restarted the daemons, then it worked.  However, Chris seems to have observed a problem with PhEdEX/FTS

The problem would probably go away even for Java on SL6 if both client and server had the update at the same time.

1. Loose ends -

	scratch space in SEs - not possible; see Sam's response before the meeting.  CASTOR can do it with its disk cache but
	then there is a copy on tape.  SEs are generally designed not to lose data if they can help it.

	storage procurement - Dan had done a relatively recent procurement and could share some info with Chris.

	information systems (again) - if publishing xroot endpoints in the BDII is necessary for DIRAC to pick them up, then it follows that
	sites that have xroot endpoints should publish them in their BDII.  In GLUE1 it'd just be another protocol (probably) and service.

	todo list for December - Elena wants to solve a space token problem. Ran the DB check script for DPM, but it will, like fsck, make
	changes to the live database so best if the system is quiescent.  It can be run in diagnostics-only mode, and may pick up some incomplete
	transactions if some are ongoing on a live system - although if there are any, those should disappear when run later.  However, to actually
	fix the issues, the system should be down.  See also chatlog.

	anything interesting you want to do in December (incl your Christmas break of course) - hmm

	anything I need to know about previous week(s) - nothing, just discussion

	updates on "xcache" - Teng reported progress at Edinburgh.  Built proxy cache; put into production.  ATLAS use RUCIO
	to find the TURL, so investigating redirecting to point to the cache.  Will try first with the second DPM.
	After this, Teng will be looking at efficiency configuration.  Once what's going on in the cluster is understood, Teng should
	be able to present/report.
	Chris has not had time to do more work but reports that it runs well in memory caching mode as reported earlier.
	Matt - no news; networking understaffed.

	scitokens and macaroons - need to look at again in December.  Macaroons are well established (a Google proposal), scitokens based on
	JWT, so IETF standard (but needs some descriptons of semantics etc.)

	on a related note AARC - and DI4R (and "data pilots") - AARC is on authentication and authorisation and included in the recent meeting
	Paul from LIGO.  Also lots of data projects involved, directly or indirectly (such as EUDAT, INDIGO DC, etc.)
	AARC proposing limited incident response testing - seems more limited than earlier ones we've done, as it's proposed to involve IdPs.
	Which is more interesting with data mover robots and how VOs are involved in the traceability.

	DPM 1.9 (did it fix the accounting problem?) - so there are two 1.9s, Matt reports, the legacy one and the bells-and-whistles one;
	and Matt upgraded to the legacy one which is "nice and boring."

2. December GDB - did they require feedback/suggestions on the access
methods?  Also the GDB has a lot of data infrastructures on the agenda
(not EUDAT though) - https://indico.cern.ch/event/578993/

   - they are not seeking input at this stage (that was easy!)  Nevertheless, for web-based access, some thought is needed to make access control
   consistent with non-web (xroot, GridFTP, SRM).  [Funny how it's going the other way - most communities are going from web to non-web, and WLCG
   the other way.]  David's hackathon is on security monitoring (SIEM)

3. AOB

   NOB



Chris Brew: (29/11/2017 10:10:50)
SOURCE [70] srm-ifce err: Communication error on send, err: [SE][Ls][] httpg://srm-cms-mss.jinr-t1.ru:8443/srm/managerv2: CGSI-gSOAP running on lcgfts08.gridpp.rl.ac.uk reports Error reading token data header: Connection closed
Java fits with dCache too
Daniel Peter Traynor: (10:13 AM)
Oye storm Se were failing ops tests with 1.88-1. they don't use argus
Elena Korolkova: (10:17 AM)
Has anyone run dpm-dbck to fix the problems (i know people run it in a dry mode)?
Chris Brew: (10:17 AM)
Steve, did you restart your Argus daemons after updating the trust anchors?
Govind: (10:24 AM)
i run it last time.. no issue.
Elena Korolkova: (10:25 AM)
Thanks, Sam
Peter Gronbech: (10:28 AM)
Kashif is planning to upgrade our DPM se to centOS7 and the latest shortly
Elena Korolkova: (10:40 AM)
sorry, need to go