Distributed transactions hard to support in MicroServices environment, strategy is 'eventual-consistency' instead
Possible future solutions
Batch Operation against multiple anchors (cm handles)
Support Yang-Patch operation for non-passthrough (cached) data
Above solutions can be added to CPS as part of a specification for London release
Michał Jagiełło support different devices (namespaces) under one root Explained Private Properties in ONAP DMI PLugin could be used to handle this with some additional modifications Not quite clear whether one cm-handle or multiple cm handles woud be used for such a device. Michal wants to await completion of multi-root functionality before considering new requirements for CPS
Kohn/London status
Branch created (for CPS Core)
User Stories can go ahead and be merged
Bugs once merged 'cherry pick' to Kohn branch as needed
it was agreed that the expected behavior is to insert prefix only when new modules occur as per the usual convention. This is a known and documented limitation in CPS and therefore not a bug. Agreed with David Terwilliger to take this as a high priority requirement for the London Release as part of a 'Feature' request from Fujitsu on CPS (the process was explained during the meeting)
it was agreed that the RFC does allow Yang modules without a revision. CPS needs to be able to accept that (handle it as an empty String) This ticket has no been converted to a bug.
Code Freeze, when can we create new branch?Team agreed to await M4 decision on because of possible issues with jobs that build release images from master branch. A Feature branch could be considered if it takes much longer.
Observation: when State goes to locked lock-reason details etc are NOT in the event Positive feedback, user story and work item related to it can now be closed
Scope of user story limited to this investigation, Lee Anjella Macabuhay wil create a follow up user story for implementation and a Spike re. the requirements for DB/Yang model updates probably without using Liquibase
Spike has proven dat model can be split without any consequences for the format of the data ie. the data is backward and forward compatible
Proposal is comprehensive but need to be converted to a CPS Wiki Page
Some functional requires separate mention, further study and separate future Jiras such as
'limit' query parameter(?) to return a limited amount of nodes
'*' representing all possible child containers or even descendants?
brackething i.e. combining several logical operators using brackets
The need/priority of cpsPath extensions should be driven by the real world need of business use cases.
any cpsPath extension need to consider the impact of the SPI (Service Provide Interface) and thus the feasibility to remain database base technology agnostic, think of an extension for regular expression support
Thanks Dylan Byrne for all your valued contributions to CPS and good luck with your future opportunities, maybe you wil find out way back to Open Source again
, and gave a list of API's which can possibly be created and updated, along with new error codes that can implemented for certain error scenarios. They will share their documentation with us.
related to ODL having transitive dependency of guava which has some vulnerabilities. Tony Finnerty suggested we should also check what version SDNC uses in order to select the correct version and still be compatible.
Questions came up during the meeting.
Rajesh kumar pointed about the response format of few of the endpoints to be in text format as opposed to json. He will reach out with consolidated details. Also there is a user story in backlog on the same lines
Ahila P pointed out that the list-node API in CPS-Core on retrieval of nodes is slower and she will come back with the sample data for the team to analyze further.
Arpit Singh wanted to confirm on resolution about a bug raised in Istanbul release related to schema-set. (we were able to fetch the details of the deleted schema-set ) which is fixed in Jakarta.
Error handling for cspPath needs to be improved. return 4xx when invalid (conform cps-core). Handle(validate) blank string separately
Inconsistencies between kebab-case for yaml modelled data and camelCase json modelled data need to be addressed. New spike
Jira
server
ONAP Jira
serverId
425b2b0a-557c-3c0c-b515-579789cceedb
key
CPS-1152
created for that
Current dmi-registry model needs to be published with note on fact that additional-properties cannot be used. Created a user story to split between public and private models:
Agree with team potential endpoints for new API in
Jira
server
ONAP Jira
serverId
425b2b0a-557c-3c0c-b515-579789cceedb
key
CPS-1133
Context: this endpoint will be an 'override' for a future more granular cache config which itself might enable dataSync (flag). The cache config will be developed using a new specialized interface
Agreed endpoint: /v1/ch/{cm-handle}/data-sync?enabled=true/false PUT/POST - No Body
if CacheEnabled is set to false after being synchronized: remove data and set datastore sync state to NONE-REQUESTED
if CacheEnabled is set to false during being synchronization: check flag AFTER sync is completed, remove data and set datastore sync state if needed
Need to enhance query capabilities, possibly model driven!
Agreed that is outside the scope of current user story. Event schema can be update later depending on what stakeholder requirements are for this. Priyank Maheshwari will record open issue on this and contact stakeholders
Renu Kumari also question use of cm-handle-id as the value for event correlation id. Team agreed to propose a separate cm-handle-id field and use correlation if (in future) when need but not for cm-handle-id. Priyank Maheshwari will record open issue on this and contact stakeholders
labels = relman AND fixversion = "Jakarta Release" AND "Epic Link" = "R10 Sign Off" AND project not in (SANDBOX, CIMAN) AND type = Task AND project = CPS ORDER BY status ASC, updated ASC
count
true
serverId
425b2b0a-557c-3c0c-b515-579789cceedb
Kohn M1 Tasks :
Jira
server
ONAP Jira
jqlQuery
labels = relman AND fixversion = "Kohn Release" AND "Epic Link" = "R11 M1 Global Requirement Approval" AND project not in (SANDBOX, CIMAN) AND type = Task AND project = CPS ORDER BY status ASC, updated ASC
count
true
serverId
425b2b0a-557c-3c0c-b515-579789cceedb
Toine Siebelink to check with Byung about need for architectural review
Sourabh Sourabh wil create a bug related to OOM depency issue for logging
Discussed Async Response wrapping. The statuscode and message in the Async response event are as intended: the original code and message from the underlying component. And NOT a HTTP code inline with the synchronous response which is generated by NCMP
stil being investigated by ERicsson Internal team. No action yet required form CPS Team
Jira
server
ONAP Jira
serverId
425b2b0a-557c-3c0c-b515-579789cceedb
key
CPS-986
some confusion about logback.xml filename still used in OOM and possible duplication. Bruno Sakoto wil work with Sourabh Sourabh to resolve this. Some additional info from Priyank Maheshwari : When a file in the classpath has one of the following names, Spring Boot will automatically load it over the default configuration: logback-spring.xml logback.xml logback-spring.groovy logback.groovy Spring recommends using the -spring variant over the plain ones whenever possible, as described here. https://www.baeldung.com/spring-boot-logging
Ahila P looked for some clarification re the integration of RAN Simulator (Honeycomb). This wil connect using SDN-C(R). SO teh full stack vor writing using Passthrough Datastore is TBDMT → CPS-NCMP → DMI-Plugin → SDN-C(R) → Ran Simulator
Toine Siebelink and Ahila also discussed the possible removal of old preloaded data. This wil be handled by a new Jira:
Solution #3, K8s is preferred by team. Is outside CPS scope i.e. in OOM or external helm charts. Document solution for now only. This does depend on new K8s release 1.20 or higher. ONAP already planned 1.21.7 for Jakarta see Database, Java, Python, Docker, Kubernetes, and Image Versions. Task/story to update OOM with startup probe once ONAP k8s has been upgraded:
Discuss with Ahila P the feasibility of supporting these items:
PATCH method support for list-nodes in CPS-Core API Ahila P wil proceed CPS Team with exact patch request CPS Team will examine possible solution but solution will be limited subset off usual Path-functionality just enough to support use-case required
Support for CPS-E-05 interface which requires Cache to be enabled In progress see Work Item #4 on CPS R11 Release Planning Cache (datastore ncmp-operational) is read-only Cache update are planned as further work items in R11 Writing to 'cache' is not planned for R11
Ran-network Configuration update using NCMP This is already supported using ncm-passtrough-operational. CPS Team will organise a demo mid next week to show passthrough read and write scenarios, see
Jira
server
ONAP Jira
serverId
425b2b0a-557c-3c0c-b515-579789cceedb
key
CPS-1004
Add CPS-TBDMT under CPS umbrella CPS Team is stil just 'hosting' CPS-TBDMT. Although it is a valuable addition to CPS it hasn't got the priority and are team is not resourced enough to take it further.
A.O.B.
Some good news Reached 1000th Jira in CPS! Great work for small team!
Some bad news Renu Kumari wil be leaving Bell Canada. CPS Team wishes her well and thanks Renu for all her valuable contributions to CPS. Renu hopes to keep contributing in a private capacity but of course cannot dedicate as much time.
