Attendees

Adam Palay (Deactivated), Andy Armstrong (Deactivated), Brian Beggs, EdwardF (Deactivated), JesseZ (Deactivated), Renzo Lucioni (Deactivated), Robert Raposa, Scott Dunn (Deactivated), Nimisha Asthagiri (Deactivated)

Meeting Notes

• Definition of Done (5mns)
  
  • Consensus on long-term goal
  • Today's Problems/Observations
    
    • Developer Velocity
      • Pull request complexity
        • i.e., over interdependencies
      • Pull request wall time for tests
      • Developer onboarding
      • Disorganization
    • Quality
      • Clear seams and contracts for tests
      • APIs
    • Extensibility/Modularity
      • Open source community
    • Incremental Upgrades (Python/Django) - Future
    • Time to value
      • brittleness
      • time to deployment
  • Question:
    • Will a containerized/modularized monolith within the same repo give us the benefits we're looking for, as measured by our metrics?
      • With the exception of:
        • incremental upgrades
        • test time - unless we come up with a smarter test-runner
        • independent scalability/failure-recovery
      • Goal for this working group: get baseline metrics, begin this work and validate/invalidate the approach.
• Metrics (30mns)
  • Before embarking on this project, what metrics can we use to validate our progress and success?
    • Collect metrics from the field
      • That extended around the monolith
        
        • Enterprise projects
        • Learner projects
      • That needed to deal with the monolith
        • Proctoring
      • That tried to modularize before working
        • Grades
        • Instructor Tasks
      • Which metrics?
        • Extensible Modularity - Repo lines touched (how many lines within the monolith needed to be changed for new features)
    • "Good metric"
      • Easy to measure
      • Actionable
        • validates/invalidates
        • not too many other contributing factors - so the metric is reliable for this project
    • Hotspots within the codebase
      • Which modules of the code have a lot of contention?
      • Number of commits touching the same file
      • EdwardF (Deactivated) to share Hotspot presentation
    • Code reviews (measuring modularity)
      • PRs that needed cross-team reviews
      • Number of reviewers that needed to be tagged
        • ?? can tag teams as well as individuals
        • Number of approvals may be a better proxy (to measure how many blockers on a PR)
      • Number of comments - (may be impacted more by other factors - such as the reviewer)
    • Pull Requests
      
      • Number of files touched
      • PRs that touched multiple modules
    • Complexity
      • Number of complexity-disabling pylint warnings
        
      • Complexity metric - trend line
    • Time-to-value (too many other contributing factors - may not be directly related to the monolith initiative)
      • Github branch age
        • Age of commits on branch - in case developers worked on it locally for long?
      • Release pipeline metrics - these are already available
        
        • Number of rollbacks
        • Disrupted deployments
    • Tests (may not be affected until tests are smarter or until code is actually pulled out)
      • Time
      • Number of flaky
• Next Steps (10mns)
  • Swarm or Divide-n-Conquer?
  • Review proposed Epics

Action Items

• EdwardF (Deactivated) Share Hotspot and Service Maturity model.
• JesseZ (Deactivated) Look into existing metric tools.
• EdwardF (Deactivated) to pull in Calen Pennington (Deactivated).
• Nimisha Asthagiri (Deactivated) Create metric tools resources wiki.
• Andy Armstrong (Deactivated) and Nimisha Asthagiri (Deactivated) to move forward on documentation.  Pull in Ned Batchelder (Deactivated).