Evaluate rollback when both old and new answers remain cached
Reverting the DNS record does not immediately send every client back to the old endpoint.
- Focused work estimate
- 5h + prerequisites
- Priority in the scenario
- High
- Engineering practice
- Distributed systems · Recovery
Estimated field mix
- Networking40%
- Distributed systems30%
- Site reliability30%
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 internal API is moving to a new endpoint. The team assumes DNS changes are immediate and has no rehearsal for stale clients.
Setup prerequisites
- Create a local resolver simulator and two loopback service endpoints with synthetic DNS records; query no external infrastructure.
Preceding work
Complete these dependencies, or supply their agreed outputs before taking this ticket.
- BDNS-101 · Inventory service records and their consumers
- BDNS-102 · Implement TTL-aware positive caching in the resolver fixture
- BDNS-103 · Model negative caching separately from empty address records
- BDNS-105 · Rehearse lowering TTL before endpoint cutover
- BDNS-106 · Preserve application compatibility across old and new addresses
- BDNS-104 · Reject alias loops and excessively deep resolution chains
- BDNS-107 · Distinguish resolver failure from application connection failure
Acceptance criteria
- Model multiple cache ages during rollback.
- Define service compatibility and overlap requirements.
- Compare DNS rollback with routing-level containment under declared constraints.
Implementation constraints
- Keep conclusions scoped to the simulated client population.
Verification to include
- Roll back with mixed cached answers.
- Expose clients that continue using the new endpoint until expiry.
Deliverables
- DNS recovery decision record.
Rollout and recovery
Retain both compatible endpoints until the modeled overlap and operational checks complete.
Value of the work
For the engineer: Practice DNS semantics, migration timing, and network diagnosis.
For the team: Produce a migration plan that makes cache delay and endpoint compatibility explicit.
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.