Make preview creation resumable after a partial provider failure
The database is created but the API startup fails; retry creates a second database and loses track of the first.
- Focused work estimate
- 4h + prerequisites
- Priority in the scenario
- High
- Engineering practice
- Idempotency · Reconciliation
Estimated field mix
- Platform engineering50%
- Cloud infrastructure30%
- Distributed systems20%
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 team creates preview stacks for pull requests. Orphan databases consume resources and previews sometimes inherit settings intended for another branch.
Setup prerequisites
- Create a local provisioning adapter with synthetic repository events.
- Use isolated disposable resources and no production credentials.
Preceding work
Complete these dependencies, or supply their agreed outputs before taking this ticket.
Acceptance criteria
- Assign deterministic resource identities per environment generation.
- Record each completed allocation before moving to the next.
- Resume from observed resources instead of blindly recreating them.
Implementation constraints
- Use a provider interface with injected failure points.
Verification to include
- Fail after database creation and resume to one database.
- Return an unexpected existing resource owner and halt without adoption.
Deliverables
- Reconciliation loop and partial-failure probe
Rollout and recovery
Canary local environments; pause allocation while retaining resource records on inconsistency.
Value of the work
For the engineer: Practice resource lifecycle, reconciliation and least-privilege setup.
For the team: Review reliable preview workflows and accountable resource cleanup.
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.