Choose timezone
Your profile timezone:
JCOP FWWG 3 October 2016: NextGen Archiver
→
Europe/Zurich
40/R-C10 (CERN)
JCOP Framework Working Group Meeting, 3 October 2017
Participants:
Peter Chochula (ALICE), Mateusz Lechman (ALICE), Ombretta Pinazza (ALICE), Luis Granado Cardoso (LHCb), Stefan Schlenker (ATLAS), Valeri Falaleev (NA62), Christophe Menzes Pires (COPMASS), Ana Sofia Nunes (COMPASS), Marc Bengulescu (BE-ICS), Pedro Agostinho (BE-ICS), Jakub Guzik (BE-ICS), Rafal Kulaga (BE-ICS), Fernando Varela-Rodriguez (BE-ICS), Piotr Golonka (BE-ICS),
WinCC OA & JCOP Framework News
- patch 20171002, based on ETM's patch P006, is now available. On Windows it features a new installer which, notably for ALICE, allows for installation into a non-standard folder.
Trending in COMPASS
- Christophe presented the concept of trending in COMPASS - see attached presentation. Each field that displays a value, or each shape representing a device has a "add to trend" option in the context menu; these are then gathered into a kind of varTrend.
- The concept seems to be very intuitive (ack'ed by NA62) and natural to use.
- At the moment ( before YETS 2017 release of Frameworks) it is unrealistic that we could reintegrate any element of this solution in fwTrending and enable it, for instance, in all JCOP panels. This should be addressed wider during the consolidation/redesign of JCOP/UNICOS Frameworks and notably their device models, faceplates, panels, etc.
- ATLAS: Some of the mechanisms, such as storing the list of DPEs to be used by the vartrend could be exposed early, so that the experiments could already integrate it.
- COMPASS raised the (already reported) ENS-17027 issue: apparently a problem with display of trend legend, where values are truncated if one uses linear axis. What they want is to be able to use "scientific" notation for legend and still use the linear axis. ATLAS suggests to verify that the "format" of the DPE (in the "common" config) is set appropriately.
followup (PG): raised the priority of this issue and spawned the FWTREND-1014 change request. - COMPASS observed that the behaviour of the Y-axis legend while zooming was improved in 3.15 (the formatting of the label adjusts itself to be well visible), but still it does not work when zoomin-out; COMPASS will open the issue with icecontrols.support to follow up with ETM
Next Generation Archiver for WinCC OA (openlab project)
- Jakub and Rafal presented the concept, architecture and the state of the project, as well as the current state of the Oracle Archiver and fwRDBAPI. See presentation attached in indico.
- With NGA there should be no need for the "RDB API" anymore (necessary mechanisms to keep the history of alias changes, etc, could be built-in); it is proposed to freeze the RDB API and deprecate/abandon it completely for the NGA
- ALICE requested clarification for long-term planning, and in particular the use of Kudu as a replacement for Oracle; should one consider such migration for LS2 and concentrate efforts on it already now?
Piotr responded that there is no hurry to do it and Oracle databases as the primary target for archiving will stay at least till LS3. In a meantime we could opt for "parallel archiving" option with data being stored into Oracle and Kudu, to gain experience and check stability. There are currently no plan to shut down Oracle service at CERN, hence it is of high importance to start the work on Oracle backend already now. Still, Kudu seems to be at the moment the most promising technology successor. - How could we sure about NGA having adequate quality during LS2, so that we could consider it for deployment (namely with Oracle backend) in some systems? Bringing RDB to a usable state took ages!
Rafal/Jakub/Piotr responded that- the architecture of NGA that separates the concerns of WinCC OA part from the DB-specific part "disentangles" one of the core problem of the RDB Archiver (monolythic code, very uneasy to debug and implement new features)
- there is a very rigorous unit/integration/system testing already now (over 450 tests already for the current prototype), and large-scale tests will follow. Quality should be assured
- by starting up the project from scratch, we do not carry-forward the technical debt of RDB archiver
- ATLAS is already using InfluxDB for their offline (probably CMS offline uses it too); it would be very interesting to see the progress on the InfluxDB backend and include it in tests. In particular the option of easy web graphing with tools such as graphana is an important advantage
- ATLAS: large scale tests, in particular for query performance should be conducted with realistic data sets (such as years of data stored by experiments). Acknowledged. We will consider how to make use of it during either the test campaign scheduled for Nov/Dec or after it
- Concerns about the lack of direct queries in the new architecture.
ETM claims that their initial tests demonstrate that the current architecture is robust enough to withstand a high load, and that the product architects strongly oppose to this element in the new architecture (additional open port per UI, etc)
Piotr clarified that the tests scheduled for Nov/Dec are planned in particular to clarify this point.
Jakub added that the message-splitting ("paging" through large result sets) that is now built-into the NGA should allow to keep the load under control - A topic about the complexity of configuration at many levels (configuration of backends thriugh the config channel, maintaining the configs for "parallel" archive groups);
Jakub explained that the configuration channel uses some generic mechanisms, yet backends would need and will have specific configuration panels.
For configs, we prefer at the current stage the flexibility over usability (the latter will be later delivered in the PARA module panels, fwConfigs and such)
A.O.B
- ALICE keeps having problem with memory leaks when fwInstallationAgent has a failed query; they tried the latest Oracle patch and it did not resolve the problem; to be followed up
- Next FWWG: likely at the beginning of November, with proposed discussion topic to be the OPC UA
There are minutes attached to this event.
Show them.