Retrospective

What went well?

  1. A working prototype ready for Control Loop demo which can act as base.
  2. Communication between components, Test bed is ready.
  3. Teams understanding on control loop design is improved when compared to initial days.
  4. Team gained fair understanding on ONAP & Policy design and test environment, tools and procedures involved.
  5. Better defined and stronger code base at present. Less subject to structural change, also easier to onboard and follow for new members.
  6. As time went on commits became smaller and more frequent. Easier to code review, merge and pay more attention to.
  7. The idea of having a Wiki Page is good
  8. Demos are really good to show progress

Where can we improve?

  1. Integration tests between different components can be started early and in parallel.
  2. Code reviews to be strict and strengthened
  3. A code review checklist can be maintained (including checkstyles, coverage, tests done, functionality covered, pending work , definition of done etc), we pushed code that did not build and did not have tests.
  4. Internal demos can be started and made more frequent, when there is a demoable component.
  5. Dedicated page of instructions for new joiners for Checkstyle/SONAR etc. Also the repositories that we need to use Nordix/Onap etc
  6. Good knowledge of Policy and the Policy Framework codebase and practices work.
  7. Add good descriptions of the reason for a commit and an explanation of the commit in comments in the review
  8. How does our architecture align with the existing CLAMP springboot approach?
  9. Understand the architectural path into the future.
  10. Meetings!

Overall:

  1. Wiki page should be updated and kept aligned before design start, as component functionality can be agreed among all. +
  2. Jira tickets can be updated with new subtasks, an acceptance criteria defined and agreed per sub-task
  3. Jira Tickets can be updated with progress and gerrit links, if acceptable.

By Component:

Participant:
1. Integration testing should be started along lines of code, and should be a daily practice instead of just relying on Junits.

Commissioning:
1.  Interaction between Commissioning and Instantiation should be defined more clearly.

Improvements

Overall:

Training of new joiners. Better information. Explain the concepts below in a good way with good diagrams and better documentation.

  • Understanding of the TOSCA service template is critical.
  • Understanding the Commissioning → Instantiation step
  • Understanding the relationship between runtime and participants and how control loops and control loop elements relate to each other.

Development Issues

PMSH Use Case

  • Describe the use case in TOSCA properties
  • Microservice, policies, CDS
  • Describe properties, setting values for properties, getting the properties to the participants

Commissioning

  1. Redefine the API's needed. Exact GET's needed + checks to introduce, What do we actually need?
  2. Pushing of control loop component definitions to participants? Flow of how you get from a commissioned control loop definition to an instantiated control loop.
  3. Checking and validation, use of reflection for properties
  4. Solve the multiple service template problem
  5. Update or deletion of commissioned control loop (name and version) is not allowed

Interaction between Commissioning and Instantiation, possibly using existing CLAMP functionality to bridge between commissioning and instantiation.

Flow:

  1. CLAMP GUI starts
  2. Get a list of commissioned control loops (REST call to Commissioning)
  3. Show a list of commissioned control loops
  4. User selects one control loop
  5. Get the control loop definition, control loop element definitions and participant definitions (REST call to commissioning)
  6. Show the GUI to get the control loop parameters (Timeout for startup), the control loop element parameters, and the participant parameters (there may not be any)
  7. User enters stuff
  8. Post the control loop instance to the server (REST to instantiation)
  9. Instantiation validate the control loop instance info (Check that references are OK and exist, check that the instance does not already exist), make sure control loop instance is in state UNINITIALISED. Also think about update of control loop instances.
  10. GUI shows SUCCESS or FAIL and error message.

Tasks:

  1. Define the Commissioning API
  2. Define the Instantiation API
  3. Examine current CLAMP GUI and see how we reuse it (Saul Gill )
  4. Implementation of the various parts

Instantiation

Interaction between Commissioning and Instantiation, possibly using existing CLAMP functionality to bridge between commissioning and instantiation.

Checking and validation

Participant types should be used in initialization, not participant instances.

Monitoring

  1. Monitoring and supervision functionality interchanged. Needs to be updated in the wiki.
  2. Need to finalize the statistics data for participants and Control loop elements. Also need to decide, if we require to implement association between statistics classes and parent.
  3. GUI for monitoring?
  4. Solve the keying problem.
  5. REST interface for the monitoring
  6. Health check (ONAP requirement)
  7. Querying by timestamp

Supervision

Refactoring of handling

Handling of error reports or missed reports from participants

Participant supervision. Timeouts, action on failure, error counts on handling.

Participants

Participant implementations for

  • DCAE
  • Policy
  • CDS

To talk to DCAE:

  1. Push Monitoring Policy into Consul
  2. Getting some information from DCAE
  3. Call a PUSH on a REST to deploy.


Refactoring Participant as a JAR

Java API

  • registerParticipant(name, participantType); Participant type matches the TOSCA participant type. eg registerParticipant("DCAE Athlone", "1.2.3", "org.....DCAEParticipant", "2.2.3"); Also could have timeout, other characteristics no of control loop elements it can run eg 20.
  • Register listener for control loop element creation. Callback when the control loop update/control loop status messages are received by the JAR.
    • Interface with two methods that need to be implemented updateReceived(updateMessage), statusChangeRecieved(Passive/Running etc)
  • Method to respond to those messages
  • Method to do stats

Test participant:

  • Allow for testing
  • Times out
  • Ignore messages
  • Block for certain amount of time
  • etc.

GUI

Rough GUI/Client

  • Supervision and Monitoring

Decisions

  1. Stick with the Wiki for documentation and convert to RST later, keep the wiki page up to date. Wiki page should be updated and kept aligned before design start, as component functionality can be agreed among all.
  2. Set up CSIT tests for the control loop work. Also get some help from experienced people in CSIT.
  3. Explain the concepts of Control Loops/Participants/Control Loop Elements in a good way with good diagrams and better documentation, also the TOSCA service template and node templates.
  4. Put up a howto on the Wiki on how to run the demo.
  5. Training on the Policy Framework and its principles.
  6. Informal Demos at the standups from everyone of commits coming in, maybe where there are issues!
  7. Get the Jenkins verify jobs running in Nordix
  8. Code Reviews
    1. Follow ONAP guidelines as a minimum Code Review
    2. Check and perform reviews
      1. First thing in the morning
      2. After Lunch
      3. Last thing in the evening
    3. Be gentle and kind
    4. Provide suggestions on changes rather than just saying something is incorrect: Do How as well as What and Why
    5. Checklist for Code Reviews (Sirisha Manchikanti to champion)
      1. Code must pass verify job, unverified jobs are not code reviewed
      2. Checkstyle: Covered by build (Verified commit proves checkstyle is OK)
      3. Coverage: 80%, Put figure from Eclipse/IntelliJ in the commit message: Investigate how to do in IntelliJ
      4. Integration Tests done statement
      5. What Functionality covered statement
      6. What Functionality is not covered (if needed)
      7. What the Definition of done for this feature is and how near we are to that
  9. Wiki page for new joiners, already is a lot of info out there but it's scattered. We should have a landing page for ourselves. Park this and come back
  10. Jira (Sirisha Manchikanti )
    1. Jira tickets can be updated with new subtasks, an acceptance criteria defined and agreed per sub-task
    2. Jira Tickets can be updated with progress and gerrit links, if acceptable.

Action Points

  1. Follow up on contract testing, also with CPS project test approach
  2. Stick with the Wiki for documentation and convert to RST later, keep the wiki page up to date. Wiki page should be updated and kept aligned before design start, as component functionality can be agreed among all.
    Responsibility: All to do
  3. Set up CSIT tests for the control loop work. Also get some help from experienced people in CSIT.
    Responsibility: To be decided
  4. Explain the concepts of Control Loops/Participants/Control Loop Elements in a good way with good diagrams and better documentation, also the TOSCA service template and node templates.
    Responsibility: Liam Fallon
  5. Put up a howto on the Wiki on how to run the demo.
    Responsibility: Liam Fallon
  6. Training on the Policy Framework and its principles.
    Responsibility: To be decided
  7. Get the Jenkins verify jobs running in Nordix
    Responsibility: Liam Fallon
  8. Create subtasks on JIRAs with Definitions of Done etc
    Responsibility: (Sirisha Manchikanti to champion)
  9. Checklist for Code Reviews
    Responsibility: (Sirisha Manchikanti to champion)
  10. Wiki page for new joiners, already is a lot of info out there but it's scattered. We should have a landing page for ourselves.
    Responsibility: Saul Gill to document your learnings on a Wiki page


  • No labels