ACTION: Update ReadtheDocs https://git.onap.org/integration/tree/docs - Jira under Integration - invite team for review - invite PTL (Morgan R.) -Gerrit PTL submission +1 from reviewers. committer/Morgan +2 (someone in Integration project will do the Merge)
New project → Repo → DOC directory (RST) in that Repo Andreas Geissler → Signoff for R6 May 2020; M0 for R7. If we are approved by TSC as a new project → create a new Repo
ONAP Architecture Review process:Project Architectural Review Requirements and Process (Draft) to M1 Project PTLs/Feature Sponsors request a review by email from the chair of the arch subcommittee. The chair of the arch subcommittee creates a JIRA issue for the review and emails a link to the project PTL. The project PTL/Feature Sponsor adds the JIRA issue link to the architecture review JIRA task in the M1 epic, as confirmation that a review has been requested.
PFB comment from TSC for the OOF SON requirement. With reference to resource commitment:
Policy: Pam - Shall I indicate AT&T shall cover it, while we subsequently discuss what all enhancements we can do in Guilin timeframe, as well as check reg. some of the aspects where you indicated use case team should contribute (e.g., separate Drools file, which would have to be done by Wipro).
SDN-R/CCSDK:Sandeep - Shall I indicate IBM?
CPS: Shankar/Ben - Given the Guilin release timeframe, I think we have to go for the simplest option (1) and indicate no impact to CPS, where we only change the yang models, and SDN-R continues to do the mapping. modify CCSDK solution to extend interfaces (OOF/SON/PCI Team updates). R7 Solution of C&PS (evolution "ConfigDB" = C&PS) Hard code information
ACTION: R7 VES Common Header update to align with 3GPP SA5 (CR) Nokia/ ATT/ Orange/ Ericsson. Presentation for Monday 2PM UTC on Alla's Req S/C call. "ONAP-ORAN Harmonization". Vimal, Marge, Cormac, Damian. Domain "Standards-Defined" NameSpace "3GPP-CM-xxxx"
The CCSDK work will be in OOF SON PCI team should be mapped to
(Generic Requirement) to consider Non-Real Time RIC Pawel Slowikowski
Model Driven C&PS PoC
Proposal (by Ciaran Johnston et al) as a step of Model Driven ONAP, Model Driven CPS. See slides in the Project Proposal page, and also in the R7 CPS Wiki.
Question: Graph DB? Answer: type of data from xNF to be hierarchical which would fit in a Graph DB. doesn't have to be that though. Likely to be one or more underlying database technologies. time series.
Question: Model driven CPS, uS on-boarding in CCSDK. If you plan to do modeling work in R7, consensus from Arch S/C. Answer: would need CPS team, modeling, Arch S/C to be aligned. Want something that can be demonstrated/delivered in R7.
Question: Are there dependencies on platform components? Answer: TBD
- Looking at internal sourcing to get development resources. Sync with theodore johnson
- Status ready to cut code off of x CC-SDK repo. Ongoing analysis. Resourcing Toine has started, 2 new people will be added in August 1, 2020. Present to Architecture S/C before end of July. Engaged Modeling S/C. Visible Yang modeled to be consumed in ONAP. Modeling work will happen in
Choice of IT DB (feeds in). Choice of DB to best fit to the use case. Implement TSDB in PostGres or MariaDB, SQL extensions for TS analysis. Issue is how good is the performance. Not as good as specialized Influx DB. Postgres 10 direct it to store TS DB similar to influx. Database derived from postgres. Timescale DB derived from postres.
Question from Samuli Silvius - State Management was interesting could we get the basic arch picture and def lines of description? Having references e.g. to A&AI is better solution than planned CPS to copy inventory data ANSWER: We are going to prepare a public page and share with you.
Network Slicing & C&PS
CPS w.r.t. Network Slicing: (Swami)
For the Network Slicing use case, we are still working out the details w.r.to RAN slicing, and are also in discussion with O-RAN folks. We expect it to take 1-2 weeks to conclude. Once it is fairly clear, we should discuss reg. the CPS impacts for Guilin.
For now, we have created Jira SDNC-1199 for the CPS requirements for this use case. As of now, we may not have resources to commit for implementing the updates in CPS (but still checking), in this scenario, there are 2 alternatives:
The impacts to CPS (which I expect to be persistence & retrieval of some basic RAN slice config related data) can be handled by existing CPS committed resource
(or)
The details are stored ‘locally’ in a DB in SDN-C for example. In the next release, it is moved to CPS.
(EMAIL) from Swami
For Network Slicing use case, for RAN Slicing, we will be needing a minimal set of information elements for initial RAN configuration and network slice-subnet specific configuration. Our current R7 GUILIN proposal is to extend ConfigDB with the required aspects for Guilin release, and once CPS takes off as an official ONAP project in HOLULU R8 H-release, the work can be taken over into CPS.
ACTION: PERFORMANCE - Open (#@#) open items to get ballpark figures for # API requests.
ACTION: TSC Scheduling & Presentation - Represent at the TSC - (TSC canceled); Tues R6 status; Thursday for general topics . Stakeholders of TSC presocialization Q&A. Schedule pre-review meeting w/ people who asked questions. SCHEDULE MAY 28 TSC 2020-05-28 and scheduled for JUNE 4 2020 TSC 2020-06-04
ACTION: Deadline? - M0/M1 project proposal objective by the TSC.
ACTION: Peer Review Process Step #1 - ONAP Projects - Presentation to ARCH S/C
Deployment in ONAP and related requirements such as documentation were agreed to be a major risk/unknown impact for this new team. Task investigating what is required need to be prioritized
Benjamin Cheung kindly agreed to present detail about ONAP documentation/process in next week meeting
Toine Siebelink will works with swam to identify E2E Network Slicing related user stories such as those required for the new models mentioned below.
Bruno Sakoto will work with Toine to define a list of user stories for the Temporal DB aspects
One all user stories have been identified a team meeting is required to agree what is feasible in the Honolulu time frame
Suggested name for API Mapp/DES like transformer: TemplateBasedDataModelTransformer
TemplateBasedDataModelTransformer will be separate service from CPS and XNF Proxy. for reason of scalability and segregation of responsibilities
E2E Network Slicing Use Case
Sandeep Shah has presented an overview 'new model' that will represent the 'external; data' not captured in the RAN Network model. Thing like Near RTRIC mapping.:
4 models are proposed, all of them contain at least E2E NW Slicing data and some also contain SON data.
Approx. 15-20 queries are being proposed. However is is expected these can all be executed using the same core CPS and xNF Proxy functionality:
qualified cpsPath query (get). assuming each model is stored/associated with its own Anchor. Or the models are tied together using an overarching parent model (yang module). To be discussed!
The goal for MVP is to deploy alternative 2 here (CPS core and NXF-REST proxy in one service) with OOM. With a possible inclusion of Temporal DB RETS interface when available (alternative 3)
Misc
The schedule for the LFN Developer & Testing Forum is now live here. Tony Finnerty& Toine Siebelink will be presenting CPS-NF-Proxy MVP on Thursday Toine Siebelink will be presenting on using Spock and Groovy for unit test
Do we need to de-scope → Agreed to flag that Update functionality might not be achieved in Honolulu time-frame.
E2E Network Slicing Use Case
Agreed name API Mapp/DES like transformer: TemplateBasedDataModelTransformer
Progress on models and query details...
E2E Slicing team will provide data sample and query information when ready
E2E Slicing team will provide query and 'update' information when ready
E2E Slicing team will provide query and 'update' information when ready
Breakdown of user stories discussion:
New function: retrieve DataNode WITHOUT children, do we need this?
Do we need separate user stories for:
Store & retrieve (keyed) Lists
Store & retrieve leaves (standard attributes)
Store & retrieve leafLists
Alternatively work on E2E Slicing models and handle 'bugs' for what is not working Team agreed to handle functionality through bugs as issues with E2E Models and Data samples arise Tests wil need to be added for specific functionality like stores leaves, leafLists etc
Update functionality (no epic or user stories yet!)
E2E network slicing only required CSP updates (no forwarding of updates to NF required as this is done elsewhere already)
What types of updates do we need (what is achievable in Honolulu) Ahila suggested to use Yang Patch payload, see https://tools.ietf.org/html/rfc8072#page-5 Will try to get more details by next week.
Demo video from LFN Developer and Testing Forum (Feb 2020)
The “knowledge master” of documentation stuff that Ben went to for help was Hagop: BOZAWGLANIAN, HAGOP hb755d@att.com
Claudio David Gasparini explained new process is ongoing. Probably need script to extend E2E use case like in demo shown today. The PNF Plug and Play Integration Test Cases Page:
Progress on models and query and update requirements
Toine Siebelink has shown overview of possible query and update scenarios. CPS Team needs input from E2E NW for the scenarios required by them with data examples and possibly SQL statements etc, Swaminathan Seetharaman will propose follow up meeting for Tuesday examples and data can hopefully be provided earlier
A.O.B. Benjamin Cheung requested CPS Team to join Use Case Realization meeting on 1400 UTC as the Intent Based Network Framework & Modeling (IBN) use case team would be interested in using CPS....
Software Status Ruslan Kashapovis finalizing study on CIST integration tests. Goal is E2E test for CPS applications. Team is free to use any testing method but might be restricted by reporting options. Infrastructure setup might suffice for H-timeframe. Toine Siebelink will contact Chaker Al-Hakimre. template for architecture is finalizing study on CIST integration tests. Goal is E2E test for CPS applications. Team is free to use any testing method but might be restricted by reporting options. Infrastructure setup might suffice for H-timeframe. will contact re. template for architecture documentation. Benjamin Cheungmade a clever suggestion to draft documentation in a Wiki and once reviewed and finalized convert to RST files. This would speed up the review process!
CPS APIsAPI methods #2 and #21 will be prioritized for H release
Both methods use E2E Slicing specific model through CPS-Core. This means NF-Proxy interface not required (for E2E Slicing)
#2 requires new query: CPS-235: Overview of (early) CPS Query query #3 This will be handled as a 'bug' as code freeze date is to close #2 also requires a 'with children' option on the query, also to be handled as a 'bug'
#21 is a simple 'get' by xpath with children which is already supported by CPS Core. Latest Model & Data sample need to be confirmed, Toine Siebelink suspect latest data.txt sample provided in is not inline with model. CPS Code has a later data sample. Niamh Corewill update the Jira with the correct sample.
TBDMP Integration as submodule to be discussed with Claudio David Gasparini TBDMP Sub repo in progress, split in separate commits: structure, code, tests. Need to think abut structure and split the PoC in required modules. E2E Slicign team will drive required tasks, track in Jira et with support from CPS Team. krishna moorthy will handle deployment? For now as uService as subschart in CPS but TBDMT should remain independent of CPS! Claudio David Gasparini can support with OOM integration
Explore Yang-Patch, E2E. Sandeep Shahshowed a basic example of a a Yang Patch with an 'edit' operation:
Patches are transactional, if one fails, all fail (and rollback of already completed steps) Patches are NOT used for adding elements to a Yang List
AOB:Benjamin Cheung invited CPS Team to join Use Case Realization meeting on1400 UTC to discuss CPS possibilities for collaboration with Intent Based Networking Project
NF Proxy Early Design, presentation by Tony FinnertyCPS-NCMP-Design.pptx Tony explained the reasoning for new terms 'Network CM Proxy' NCMP 'CMHandle' which will replace the term 'Anchor' in the context of CM Proxy recommend to read about NMDA https://tools.ietf.org/html/rfc8342
M3 :'code freeze' Thursday
PLEASE CLOSE ANY TASK THAT IS DONE! Change 'fix version' tag to Istanbul for anything that wont (need to) make Honolulu
E2E Network Slicing Use Case
CPS APIsAPI methods #2 and #21 will be prioritized for H release
Build issues. Claudio David Gaspariniexplained we need to raise a ticket on LNF and create a ci-management change.
OOM integration, we are probably too late for this but krishna moorthywill start process anyway with support from Bruno Sakotoand Claudio David Gasparini OOM integration of TBDMT needs to remain independent of OOM integration of CPS! As a fall back option TBDMT can be manually added to an ONAP deployment for demo or a local deployment an be used.
Need to support at least 2 CPS endpoints.. Get and 'Query'. The query endpoint is ' <server>/cps/api/v1/dataspaces/dataspace/anchors/anchor/nodes/query Demo available in this video. Krishna wil handle this requirement as a 'bug' after M3
Swaminathan Seetharaman will create new REQ's for E2E Slicing and SON PIC (scope to be still discussed internally)
Future of TBDMT, not clear is nice as part of CPS but will require additional resourcing as much work still needed like integration, conform to CPS-Core quality standards etc. Swaminathan Seetharamanand krishna moorthywill prepare slides to follow up this discussion next week with a view to take it to Arch. Committee.
E2E Network Slicing Use Case
CPS APIs How to continue? Proposed to split according to 'operation' 2 teams of 2. Toine Siebelinkwill create Jira's to track this work. If needed CPS team can meet E2E team middle of next week for further clarification. see and
OOM blocked, RC0 postponed by 1 week CPS team just waiting on OOM to merge (gating jobs are passing) CPS-290 more scenarios of 500 error to be discussed
Presentation on E2E slicing Check Andy MAYER (ATT) Ruhal Tyagi (Ericsson)
'Home' for TBDMT for review with ArchCom Template need to be aware of Model/Version Versioning could be part of ID Template should be applicable to range of versions Preferred by Team: TBDMT stay in CPS. Main concern is Resourcing. Swami will add slide with options and Pros & Cons and review slide again next week
Owner of slicing 'config' data has to be the 'use-case' (not the TBDMT)
Process for dopping back Toine Siebelink will document a proposed WoW for bugs on confluence with help from team In general all bugs should eb fixed on main branch and dropped back to previous release. Consider API changes, user story Acceptance Criteria etc. If there is no API/behavior impact and it didn't break acceptance criteria then it could be an 'enhancement'. Claudio David Gasparini will prepare for discussion on backward (in)compatible updates next week
AOB
Agreed to change meeting duration back to 1h starting next week
Please note DST has started in US, This will affect attendees in zones with different DST rules
- Official waiver from TSC for so it can be postponed to Istanbul (no need to dropback to H) - Security Hotspot, needs to be flagged as False Positive. Amy Zwarico will follow up on this
Only impacts CPS (no other ONAP components) so now classified as a 'Specification'
SDNC (integration) Using RESTCONF ODL Interface Test Only impacts
Claudio David Gasparini suggested to use Java-API client? Would be significant arch. change! Take over responsibility from SDNC, breaking µS concept, scaling concerns etc. Maybe consider/discuss this again later e.g in J-release
Preferred option for Model ingestion: direct from xNF (based on change event from A&AI)
Question on SON Use Cases: Model information is information, Niamh Core will send an email. Also please confirm SON use case is stil considered for I timeframe.
CPS Team need SON Model, awaiting response from Sandeep
Need alternative solution but Renu Kumari will comment and propose way forward Regarding OOM Renu will also create a general OOM-updates ticket based on a pattern used in 'Policies'. Any OOM updates can be added as task to that ticket. Traditionally this was handled at end of any release but we are awaiting ng an OOM process improvement to have an earlier staging release to make these changes. Renu will communicate to team and OOM about this process
Discussion with E2E Network Slicing regarding 'hardcoded ran-network-model':
Swaminathan Seetharaman (or his team) will investigate the feasibility of model-retrieval using RESTConf and NetConf from the Honeycomb(?) simulator used for their use-cases. Following the produce investigated and documented by Rahul Tyagiin
A&AI does currently not include/cover the simulated nodes used by E2E NW Slicing. NCMP should NOT try to solve this using SDNC/R or similar. If needed a 'shortcut' using a hardcoded list(file) with the simulator data until this functionality is added to A&AI.
A dedicated ONAP-Honeycomb-DMI plugin could be considered if needed e.g. if models can not be retrieved over RESTConf.
At the same time it might be required that NCMP in the I timeframe support both simulated nodes and nodes available in A&AI. Prioritization and feasibility (depending on resourcing too) of these options need to be discussed
At boot time of DMI Plugin it will need to inform NCMP (DMI-Registry) of all CMHandles its manages
Spikes by Bruno and Renu to be reviewed by remainder of team (follow up meeting early next week)
CPS-36. Rishi remarked there is no docker-file yet for TempDB. Wil create separate task to do that first
Discussion on the fact TempDB is juts one module. It was greed that for simplicity it wil be kept like that until the need arises to split it. Bruno is also introducing an Architecture Quality Tool to ensure the singel module still has a god architecture that can always be dived into modules in the future
TBDMT. Question arose around need for (read-the-docs) documentation etc. Toine Siebelink will the requirement for TBDMT and make sure documentation is part of that.
New tasks suggested
Update read-the-docs for CPS with high level description of its components
Update & reorg wiki with good overview of current architecture of CPS and its components
Note. Pass-thru wil include write operations. In Istanbul release only full re-sync wil be available to update CPS Core Cache after write operations...
E2E Network Slicing Update
Simulator findings, not available yet but Honeycomb is intended to be fully NetConf compatible. It might mean impacts on simulator are required...
Short discussion re. SDNC & SDNR Tony Finnerty & Swaminathan Seetharaman SDN Controller is Prime interface for CM to Network. Owns the interfaces to network. NCMP owns the cached data. NCMP that requests models & data from controller as needed.
Renu Kumari: ONPA want us to use authorize docker hub access for Postgres images. Create to update CPS-temporal, will create a similar user story for CPS-Core
Create story for update CPS-Core REST with output description and samples (done)
Model operations are NOT part of the interface discussed. But a (Separate) interface for this is planned and minimal support in Istanbul timeframe is required
NCMP wil use different datastores (dataspaces) for the same and/or slightly different models for the same xNFs. This might requires soem attention/future optimization
Possible on SonarQube coverage reporting in CPS-Temporal . Bruno Sakotowill create bug to address this
CPS is part of Trial for new way working for SonarQube:
ONAP Way profile created (copy of Sonar Way) which gives us control of when SonarQube updates come in
PTL (Toine) has now capability to flag false positives and mark Security hotspots as 'safe'.
Discussion regarding exact scope Istanbul
E2E Network Slicing detailed requirements
Will use CPS-Core only.
Upload model using rest endpoint as needed
CRUD Operations against data in CPS-Core only for now
TBDMT will have some minor impacts when switching to NCMP interface at a later stage (ie. specify dataspace). In any case it seems required that TBDMT can interface with both CPS(core) and NCMP
Ownership of data is important but does not require impacts in Istanbul
Update on Simulator investigation
Honeycomb RAN Simulator should have capacity to retrieve model over NetConf/RESTConf interfaces using ODL
some open issues left re. datastore names and (fallback) behavior
REST URI format agreed
Ahila P Some clarification required re. DMI-Plugin, node, model and data discovery
Step 1 Node Discovery - If E2E simulation node not in A&AI: DMI PLugin has side-door REST endpoint to add (initial) inventory
Step 2 Model Discovery After node is Added, NCMP wil trigger Model Discovery when needed 2a NCMP wil request list of module (name)s from Node 2b NCMP wil request yang-source for any module sit hasn't stored already (for other nodes)
Junfeng Wang has some concerns about pre-conditions for node discovery. Ahila will add these details later
DMI-Plugin interface, abstraction that allows flexibility to have other ways of discovering initial inventory and inventory changes. Intention is that ONAP will only have one DMI-Plugin. Using a temporary 'side-door' to add any cm-handle not available through A&AI
Step 3 Data Sync Not worked out yet Similar process and component involvement as Model Sync
Bug Ruslan Kashapovhas attached the correct model. Is compatible with previous data already in system
'Bug' is not a bug but a task or user story that needs to be executed as part of E2E slicing use case support
Latest findings on config=true/false support using SDN-C/ODL by Rishi Chail: ODL supports 'content' parameter and translates it to the equivalent NetConf operation(s). However current the tests how the same output even tough some elements are flagged as config 'false'. Claudio David Gasparini suggests using a different simulator to identify where the problem is. is and wil help Rishi with this. Oskar Malm noted that most current nodes do not support NMDA datastores (yet) So intention for Istanbul is to only support non-NMDA devices. Non-NMDA device will also use one model for config-false and config-true data.
Michael Kerrigan Build problems, need to check what test is failing exactly
Renu KumariNeed agreement on Kafka default authentication options and population of related env. variables. Agreed to 'follow' CPS-Core but further discussion needed with Bruno and Aditya. Docker compose setup for Kafka, Agreed to handle taht as separate task for CIST testing. Renu Kumari will create ticket for that
Dylan Byrneteam agreed change of 'additionalProperties' to 'cmHandleProperties' as the former is an Open API reserved word. Niamh Core wil update Wiki pages affected.
was explained and has detailed steps and data to reproduce. Will be assigned to CPS Team member shortly
Toine Siebelink has sent Ahila Pdocumentation on how to use text() function. There was also a mismatch in property-name. Ahila will try again.
Add list element and updating properties wil have to be done in 2 separate calls! Toine Siebelink has sent Ahila details about CPS-Core Patch URL to update leaves only.
Holiday Cover
Bruno Sakotohas been set up for LastPass but need to test this. Bruno will be absent too. So team has to plan for this..
Shiva Subedi started presenting CPS-506: List all known modules and revision team accepted reasoning to only use module-name and revision to uniquely identify a module. Namespace is not needed! has been created to deal with incorrect modules at some point in the future.
Bug in progress by Dylan Byrne. Also relates to an issue regarding additional properties not being saved in
Discussed workshop held by Ericsson team in Ireland, along with sprint retrospective.
Brought up current issues with code reviews, and how they can be conducted in a more efficient manner going forward, by breaking down commits into smaller chunks, and not recommitting non blocking issues where necessary approvals have already been made.
Renu Kumari brought up the question; how will these further commits be documented? And this was also discussed by the team
Dylan Byrne will create a code review best practices wiki which the team can contribute to with their opinions on what should count/not count towards blocking issues.
Many SQ violation in CPS-Core and NCMP mainly about var-rule. Please use SQ locally to prevent these. Toine Siebelink will also discuss the value of the rule in PTL an see if it can be disabled
TBDMT Quality is also regressing and improvements proposed for Istanbul to allign with CPS standards haven't been addressed yet.
Integers are represented as floats ('.0' added in output) when read from DB. CPS Team will investigate now the issue is understood better
Issue seems in TBDMT, awaiting response from krishna moorthy
Logging. Lack of logging Toine Siebelinkhas created task to document capture of CPS logs: It might lead to an improvement to increase logging on positive scenarios.
Code Freeze & RC(0)
Team will need to focus on bugs, doc, CSIT & Demos. CSIT test and Demos might possibly be combined. Includes but not limited to:
Release Process & Bugs. Bruno Sakotoexplained the process
