JCOP FWWG 4 Oct 2022
JCOP Framework Working Group Meeting
4 October 2022
Participants: Piero Giorgio Verdini, Christophe Menezes Pires, Luis Granado Cardoso, Luis Miguens Fernandez, Ivan Atanasov, Clara Gaspar, Mateusz Lechman, Piotr Golonka, Cristina Vazquez, Frank Glege, Ben Farnham, Juan Manuel Pablo Rodriguez, Piotr Nikiel, Jean-Charles Tournier, Peter Bond, Juska Pekkanen, Stefan Schlenker, Geraldine Thomas, Michael Ludwig, Mariusz Suder
INDICO link (with slides): https://indico.cern.ch/event/1203629/
Discussion of the CAN->SNMP migration for the Wiener component
The topic was presented by Michael (see the presentation attached). The following comments have been raised during/after the presentation
- the motivation for this migration is that actually one cannot purchase CAN-based fantrays from Wiener anymore! ATLAS have a couple of them to be replaced after a failure and need to use the SNMP ones
- clarification: the one and same Wiener OPC-UA Server is used for both the CAN and SNMP based hardware; OPC items names obviously differ in the structure
- Q: how is the migration in the approach "1" going to be implemented? A: migration panel that would act on the address config of existing datapoints (without renaming them). A prototype of this panel already exists
- Q: how many channels are affected e.g. in ATLAS? A: currently there is a immediate need to convert 2 fantrays. Then progressively more and more could be envisaged. One would expect of the order of thousand(s) of channels.
- Take note that in scenarios other than 1 (or 3) ConfigurationDB recipes would need to be re-done. Static configurations will need to be saved (with updated addresses) anyway
- For option 3, with duplicated datapoints, the output address could be set on only one of the datapoints (otherwise you get warnings), which could be a strong limitation. Is it for the new DPs only then?
- There should be a rollback procedure in the migration panel
- reapplying the default framework addresses to the devices with standard settings could be used
- ASCII Export of the datapoints before the conversion could be an option?
- LHCb would definitely support option 1 as it is securing the compatibility, ie. for cases such as panels/scripts written by subdetectors, etc
- CMS: would also favour option 1, yet with a possibility to rename the datapoints, such as "can0" could be renamed eg. to "snmp0", etc.
- for CMS this should not lead to the loss of archive
- for others, the impact of such dpRename needs to be reviewed: most likely it means loss of access to archived data - to be clarified with @rakulaga what wlill be the effect
- still, an option for doing this should be provided for the use by CMS
- The conversion panel should allow to select also the driver and the non-default OPC Connection that would be used. This is similar to the OPC DA->UA migration panels.
- the above functionality is also covered by the fwOpcUaTools component....
- it would be nice to have it combined to the migration panel though
- it will require a refactoring of fwOpcUaTools to move certain functionality from a panel ("tool") to the library where it could be reused, etc.
- Q: What will happen when re-discovering the hardware in the OPC UA Server? Are the new names "predictable" (pre-determined)? The names (i.e. at the top of the root of device hierarchy, the fantray name) is fully configurable in the discovery-configuration file; even with a bare minimum you would define the names and the IP addresses, and they will then be used in the generated config files
- Q: What if we had 2 crates on 2 different canbuses having the same name?
- We would need to prefix them somehow... For this reason the conversion panel could have an option to edit the "prefix" (the "root" of the hierarchy), so that one could have eg Wiener/can0/myCrate1 -> Wiener/snmp0myCrate1 etc. This would mean a dpRename would be needed - see the comment above about impact on the archive
- In any case, once we have such a rename, it will be again possible to have the OPC-UA Server config file regenerated from WinCC OA. Resetting device addressing to defaults could also work (provided that the default driver and OPC connection is still in use)
- If we wouldn't do the renaming, then the reset of addresses to defaults will certainly mean generating wrong addresses and loss of communication. To be handled with care then!
- During the conversion it should be possible to customize the top of address name (ie. the fantray name) that would be used for configuring of addresses, without renaming the DPs. In other words: we should not blindly convert can0.MyCrate1.XYZ to MyCrate1.XYZ but have the option to change it to e.g. Snmp0MyCrate1.XYZ
- Option 3 would certainly require moving to another driver/connection
Comment sent by Cristina after the meeting
I was revising fwWiener/fwOpcUATools and actually there’s no automated process to produce the configuration file for OPC UA (there wasn’t either for OPC DA). Users had to produce it on their own.
I was mistaken by fwElmb which actually has it and by our CMSfw tools that we developed for migration to OPC UA when fwOPCUAtools was not yet available which also allows to produce this automatically…
So actually, if you want to fully automate CAN to SNMP migration, you would need to create from scratch some tool to produce the server basic configuration file. The most straightforward is to do this based on the info from fwWiener crate DPs. Another approach would be to leave this to the user, they would just need to recreate the configuration file just as they did when they migrated to UA initially or every time they create a new Wiener based system. In CMS this is no problem, we have already automated tools to do this during Wiener based component postinstall or through a panel (it is nevertheless following some CMSfw name conventions for the server connections and OPC UA run directories).
In summary, consensus has been reached that the option 1 (ie. keeping the datapoints and re-configuring their peripheral addresses) would be the most suitable. However, a number of features discussed above should be made available in the migration panel
- The possibility to choose the driver and OPC Server for the converted addresses (ie. others than default ones)
- The possibility to change the "topmost" part of the address, ie. set a custom name for the fantray; this should be in addition to the default behaviour where canX.MyCrateY.SomeItemZ would be shortened by simply removing the canX part
- The possibility to rename the devices (without a change of DPIDs) using dpRename to set a new prefix (notably for CMS)
- Implement rollback scenarios in the panel
Update on the "ideal Can Gateway"
The prototype hardware has arrived. First pictures of it attached in INDICO
Concepts and model for the new Trending tool
Presented by Mariusz - see the slides that are attached.
- Q: Wouldn't it be better to implement data sources in something more than CTRL and something else than "direct connect"? How about embedding eg. Grafana in an EWO?
- A: Some users would still prefer to use the visualisation based on the trending widget, notably due to the fact that it is much more lightweight.
- LHCb stronly supported the above statement
- Clarified that these are distinctive use cases that could be looked at independently; each (embedding the dashboard vs using trend widget) having its own advantages
- LHCb commented it is looking forward to see the final version of the tool.
AOB
- No plans yet for the next JCOP FWWG Meeting at the moment. In November we expect the presentation of results of the study of performance/scalability of database technologies to drive the choice of one that could be used with NextGen Archiver
- Beta release of WinCC OA 3.19 expected in around 2-3 weeks
- CERN yearly followup visit to ETM planned for 6-7 December. Please, signal through your Controls Coordinators any topic you would like to have raised.