AWS Site Reliability Engineer Interview Questions

I would identify the critical dependencies and decide what can be cached, degraded, retried, or isolated. Then I would design timeouts, fallbacks, and circuit breakers so one dependency cannot take down the whole service. The reason I choose this approach is that resilience comes from planning for partial failure, not hoping dependencies always work. I would also think about queueing, bulkheads, and whether the service can continue in a limited mode. If the interviewer wants more detail, I would connect the design to a specific production failure scenario.

I would define the key SLIs for each service, then build metrics, logs, and traces around them. The monitoring system should let engineers see both high-level service health and request-level failures quickly. The reason I choose this structure is that a large microservices environment becomes hard to debug without good observability. I would also think about dashboards, alert routing, service ownership, and how to keep the signal useful rather than noisy. If the interviewer wants more detail, I would explain how I would validate that the monitoring actually catches meaningful incidents.

I would use a backward-compatible migration plan so old and new versions can run together during rollout. A common path is to add the new schema, deploy code that can read both versions if needed, backfill data, switch reads, and then clean up the old path later. The reason I choose this approach is that it lowers downtime risk and makes rollback safer. I would also think about locking, migration size, and how to test the migration on staging data. If the interviewer wants more detail, I would discuss how I would handle partial failures during the migration.

I would first clarify the consistency requirements, because active-active designs trade off complexity against resilience. Then I would design routing, replication, conflict handling, and failover so traffic can be served from multiple regions. The reason I choose this approach is that active-active can improve availability but it needs careful handling of data consistency and user experience. I would also think about how to keep latency acceptable across regions and how to prevent split-brain behavior. If the interviewer wants more detail, I would describe how I would test regional failure scenarios.

I would collect logs from all services into a central pipeline with consistent formats and correlation IDs. Then I would index and store the logs in a way that supports search, filtering, and retention policies. The reason I choose this approach is that centralized logs make incident investigation much faster across many systems. I would also think about sampling, storage cost, and access control because logs can grow quickly and contain sensitive data. If the interviewer wants more detail, I would discuss how I would balance observability with cost.