Lukasz Rajewski Discuss use of cpsPath for get operations (instead of absolute xpath support only)
Toine Siebelink explain the core reasons for NOT combining these
Architecture design to separate 'get' from 'query' functionality: Client should be aware what they use when
Although xPath and cpsPath can look very similar the cost and complexity of processing a cpsPath is order of magnitude higher
xpath cannot easily be distinguished from valid cpsPath unless the model is parsed too which again would lead to additional cost to support both on the same interface.
Operation that modify the data will only support xPath too. This is again an architecture decision that is based on best practice, scalability and transparency. If the get operations would support cpsPath the same would be expected for the operations that modify data too and would violate the aforementioned principles.
Delta interesting but prefer to use dataspace/datatores in future for this instead of CPS Temporal
Support with disabling notifications dataspace, details to be forwarded to Arne Chrestin
Ahila P write operations performance was not prioritized in London release instead the focus was on read operations and some queries. In Montreal we wil shift toward optimizing write operations too. But the team is short on resources and have a large backlog. David Terwilliger will try to get additional resourcing for the CPS team. Ahila will create requirement Jira for optimizing write operations.
Ahila P Pagination wil be progressed by DT/Tech Mahindra (also need tracking JIra for R13/Montreal)
Toine Siebelink discussed this with Tony Finnerty and agreed CPS should be abel to handle DB down and have a retry mechanism too. So apart from ETH making deployment changes CPS team should also provide a build in robustness on DB connection...
Introducing Anirban Bhattacharjee Product Owner for CPS
CI (docs) Build issue status
Ongoing CI Team aware troubleshooting ongoing
CPS Team can IGNORE this (new) job for now as long as all other jobs pass any Comitter can delete the -1 from CI Build and set Verify to +1 instead if need to merge pending commits
Scalability, has been test CPS Core and scale using multiple instances but quickly DB (Postgess becomes bottleneck) CPS current focus on optimizing code to get maximum out of singel CPS instances. There is scope here and good improvements are already seen.
Two top element trees, patch using xpath /, error reported. Although support would be 'nice' there is no urgency. Need to check with DT stakeholder re. priority
Two top elements but no children just leafs, works partially (only one tree would be updated,a s expected). Expecting such simple model is unlikely and fixing 1. above should fix this too
List example problem is unclear, needs to be updated. It seems unrelated to multiple root scenario and possible legacy issue(bug). Arpit Singh will separate this into a bug on its own.
model change required. Lee Anjella Macabuhay has tried to organize meeting with Ahilla and Robert Vagra but Robert is hard to get. In any case Andras has been able to use the given model with the recommended fix in the latest version of CPS to do further performance testing as part of
Jira
server
ONAP Jira
serverId
425b2b0a-557c-3c0c-b515-579789cceedb
key
CPS-1475
Toine Siebelink demonstrated in a bit mode detail the instrumentation introduced by
Arne Chrestin has provide small fix and CPS team wil review and merge later today if possible.
Ahila P reported some progress on performance testing with some possible changes. Toine Siebelink will call meeting early next week to coordinate her and CPS team's efforts
Depth functionality works well, some updates required to OpenAPI documentation: spelling, use " quotes for text values, explain 0 and positive values, only mention -1 as a note.
Ahila P believes it is a ODL parser issue and that the model used to parse with teh previous version of ODL. Lee-Anjelle will double check and then organise a follow up meeting
Lee Anjella Macabuhay ODL 8 upgraded completed. ODL 9 requires Java 17 (are on 11) Need to investigate...
RTDv3 CI Jobs are failing this is a know issue . CSIT (in CI) should be OK. If for a given commit only the RTDv jobs fails and there are no RTD changes that commit can be approved (Toine Siebelink can set the verify flag +1 if needed). Toint to check teh status of the related fix with LNF support: Matthew Watkins
Levente Csanyi reported issues with the performance tests in Nordix CI. It might just be the time limits are to close and teh variation is too large. Levent will check
API Version handling proposal OpenApi / Swagger UI
apiVersion for any method that support both versions
hardcode v1 or v2 when needed
Output imitation using 'depth' parameter
agreed on v2 version of method (document affected methods) with just 1 parameter: 'descendants' (type string). Suggested acceptable values 'true',' false', 'all', 'nono' (case insensitive) number '0' and any positive integer value
Toine has applied workaround that makes Robot CSIT tets working again, see https://gerrit.onap.org/r/q/topic:%22RELENG-4523%22 but his has to be undone when RELENG tiket is properly resolved. But the team is unblocked now
A.O.B.
rajesh kumar looking for update on requirements from Ahila P : Long term yes but nothing concrete, but would like to join efforts. Consider just using Depth-parameter for now as already internally supported buy CPS service. Rajesh wil start investigating and create ticket.
XML Issue: data-wrapper required by XML. Will try to Poc Wrap/Unwrap with data as needed, communicate with ODL community about ODL support to automatically wrap it as part of parsing
Jakarta fix, build issue Luke Gleeson ssuspec CSIT depend on a now deleted image. Will contact Michal next week.
Andras Zoltan Kiss and Levente Csanyi still having issues deploying SDNC because of the amount of dependencies. Toine will call meeting to follow up (called for )
Ahila P asked for plans re. 'Pagination of large output'. Toine Siebelink explained there are no plans and expressed concerns because of the nature of Graph Data. Support of 'depth' parameter already available in Java API. Support could easily be extended to REST API. Tony Finnerty mentioned that Netflix might have solution in the open source domain. Michał Jagiełło suggested to check with A&AI since they have graph data and some form of pagenination implemented
Ahila P explained their requirements for prefixes and showed some preliminary code changes. Also indicated they need a 'short-term' fix before a a ore costly 'long term' solution
Toine Siebelink will call meeting with interested parties next week to discuss the issues, ensure agreement in communities and see how best to proceed implement a short term and a long term solution
