Aglet

Investigate whether a rollback changed the served version

The investigation should resolve the gap between rollback intent and observed serving state. Follow immutable release identity through deployment, routing, runtime, queue, and customer behavior records, retaining delayed or conflicting evidence as uncertainty.

Build a useful investigation brief

  1. Build a version timeline

    Order original release, rollback decision, rollback artifact, status transitions, routing changes, runtime observations, and symptom records. Mark timestamp precision and processing lag so a sequence is not mistaken for causal order.

  2. Trace exposure boundaries

    Compare environment, route, cohort, worker, cache, and queue versions before and after the rollback. Identify any participant whose serving state is inferred rather than observed.

  3. Compare behavior evidence

    Use representative failures and successful completions to test whether the original path changed after the rollback. Check traffic volume, retries, and delayed effects before attributing improvement to the rollback.

What to carry forward

The brief should state the strongest evidence for served version, exposure exceptions, behavior change, and competing explanations. Conclude confirmed rollback effect, partial containment, reporting-only change, or unresolved state with confidence and next evidence. Trace rollback intent, status transitions, routing, runtime identifiers, and customer behavior in order, marking delayed records and replacement releases instead of assuming the first status is authoritative.

Keep the decision with the work.

Use a Work Item in Aglet to record the problem, the evidence you have, and the next decision. Add an owner and priority, then keep updates in the discussion so the next person can follow the reasoning.

Create an account See the product workflow