NG Tech meeting, NG2017, Tromso. 9:15 - 18:30 

breaks: 10-10:30, 13:00-14:00, 15:40-16:00


Present:

Ulf, Jozef, David, Maiken, Roger, Dyma, Martin, Christian, Aleksandr, Mattias, Florido, Andrii, Gianfranco, Farid, Andrej,
Mattias  Ellert, Cedric, Boris, Oxana, Vincent, Balazs (notes)  


0) 9:00 - 9:30 welcome, agenda 

1) 9:35 - 10:10, Maiken  Status Report from the ARC release manager: release plans, release procedures, svn status. 
Content of the next major release.

GIT discussion.

ACTIONS: 
-Maiken prepares git-usage page (initial ideas)
-Maiken extend release procedures to cover testing
-Maiken to initiate Naming/versioning discussion 

next major release: emphasize the improved Vo support.


2) 10:15 11:05 - Martin,  Client-side development: API status,  support status and plans for python, java, etc..  and the bad practice of coding against the CLI. 

ACTION: fix SDK documentation, implement all the code cleanup & removal decisions. For the latter we need a dedicated skype call.
ACTION: Investigate GUI status 
ACTION: investigate python problems reported LHCB
ACTION: Andrej posts a vision draft for the ARC client
ACTION: Martin goes through the open client bugs and prioritizes them. 

DECISION: we abandon JAVA, best effort, no support. JAVA will be dropped from the next major release

CLI: important for Slovenian community. dozen couple of Finnish users, sysadmins, developers 


11:10 - 11:30 cofee 

3) 11:30 - 12:20 Maiken, aCT @site
ACTION: splitting internal A-REX control directory handling as a separate unit/module. 
ACTION: update figure, find proper name for the new plugin (no local please), make it better sale-able (marketing wise)


4) 12:20 - 13:15 Andrii,  The next generation Information Index Service for ARC: ARCHERY, the ARC Hierarchical Endpoints Registry. 
--Discussion on deployment, operation, migration & retirement of the old EGIS framework.

ACTION: Andrii post proposed deployment scenario for current nordugrid sites & services (top indices, country indices, atlas index).
several proof of concepts: everything in one index
ACTION: Extend ARCHERY to contain all service endoints, including job management too! At the same time drop EGIS services from ARCHERY records.


13:15 - 14:05 lunch

5) 14:10 - 14:30  Mattias Ellert, Living in a world of no Globus: the effects of the Globus support discontinuation news. 

ACTION: approach condor and ask them again to support ARC submission via EMI-ES now that globus is obsolete

ACTION: test xrootd, https data transfers in ARC. Switch the production NDGF storage element NOW.   

note: LCAS, LCMAPS will be a problem if globus dissappears


6) 14:30 - 15:10 ) Vincent,  Evolution of ATLAS Distributed Data Management and its possible impact on ARC future, 


ACTION: investigate and demonstrate howto use ARC as an unmanaged storage


7) 15:15 15:33 - Dmytro, Elasticsearch and ARC


ACTION: Dmytro will use logstash to upload metric files to elastic search. 


15:35 -  16:00 cofee


8) 16:00 - 45 Balazs, The new ARC config (server-side). 

ACTION: Balazs sends seleton config to Dyma who comes with a realistic set of values for a minimalistic cluster config 

ACTION: define which blocks are mandatory


9 ) 16:45 - 16:55 Andrii, overview of the new  parser

ACTION: expose full dictionary in python 


10) 17:00 - 17:35  Oxana, VO concept :  Managing VO-based access control in ARC (authgroups and other VO related concepts) versus VO's in ATLAS/WLCG. 

ACTION: come with a dedicated VO-centric config block where a single block is used to list all the queues, services, mapping info for a specific VO. 
 

 11) 17:40 - 18:05 Ulf,  Misuse of mutexes and broken state of locking in ARC. Problems with sqlite3 delegation db. 
 
locking:  waiting for support for newer c++ versions on all platforms
 
sqlite: ACTION: Ulf comes with a better schema


12) 18:10 -  19:30 Aleksandr, Post mortem analysis and report on recent ARC instability troubles: the problem(s), the fixes and the lesson learnt.  

ACTION: define heavy load tests. 

ACTION: continue with the collection of logs & performance data
 
ACTION: there is still some instability, investigate further (see Maiken's ganglia chart)
 
--------------------------------------------------

postponed for next meeting:

x) Infosys I/O performance tests in the control-directory. 

x) LRMS python lrms status

x) anoymous monitoring via EMI-ES interface