A deployment causes intermittent 500s. What do I check?

Short explanation:

Intermittent failures after deployment often mean only some instances, requests or data shapes are incompatible with the change. A rolling deployment can briefly run old and new versions together. Grouping errors by instance and version reveals whether the issue is configuration, code, readiness or schema compatibility.

Key points:

• Group errors by version, instance and route; inspect readiness and config.

• Check compatibility with database changes and old instances still serving traffic.

• Roll back safely if needed, then verify error rate and key user flows.

• Check migration timing and whether old code can still read records written by new code.

• After rollback or repair, verify critical flows and new background-job formats as well as the HTTP error rate.

Interview point:

• A rollback is safe only if the old artifact can operate with the current data and infrastructure state.