Canary one cohort with a measurable abort rule
A canary is called healthy because no one watched it for long, despite a synthetic error spike in one endpoint.
- Focused work estimate
- 4h + prerequisites
- Priority in the scenario
- High
- Engineering practice
- Progressive delivery · Release analysis
Estimated field mix
- DevOps60%
- Site reliability40%
Field percentages are editorial estimates of the ticket's engineering focus. They total 100%; they are not measured time, proficiency scores, or ownership evidence.
Review it, then add it to your workspace.
The board opens an editable draft; nothing is saved until you confirm it. Sign-in and workspace permissions apply, and Demo boards remain ephemeral.
Project context
A fictional scheduling API is rebuilt separately for test, staging, and production-like local environments. The resulting images differ, release notes list the branch instead of the artifact, and rollback rebuilds old source with new dependencies. Use a local registry simulator and synthetic deployments; no cluster or customer traffic is supplied.
Setup prerequisites
- Artifact digests
- Release states
- Health checks
Preceding work
Complete these dependencies, or supply their agreed outputs before taking this ticket.
Acceptance criteria
- Canary binds cohort, duration, traffic minimum, and comparison baseline
- Abort evaluates declared error and latency thresholds
- Insufficient observations remain inconclusive
Implementation constraints
- Use deterministic synthetic traffic; thresholds apply only to this exercise.
Verification to include
- Run a healthy canary through the full observation window.
- Inject an endpoint-specific regression and confirm automatic halt before wider rollout.
Deliverables
- Canary policy, evaluator, and regression fixtures
Rollout and recovery
Expand only after a conclusive passing result.
Value of the work
For the engineer: Practice immutable promotion, release authority, compatibility gates, canaries, and rollback.
For the team: Review a delivery flow that can identify exactly what is running and restore a known artifact without rebuilding it.
Evidence boundaries
Outcome Evidence: Tests, patches, and runbooks are requested deliverables. They become Outcome Evidence only through a qualified Mission and immutable Evidence IDs.
Ownership Evidence: Independent adaptation must be observed under a declared verification policy and cite immutable Evidence IDs. Completing a planning ticket establishes no Ownership Evidence.