Loading…
Operational design interview questions for SRE, DevOps and platform roles: deployment pipelines, rollout and rollback, health checking, secrets rotation, config and backups.
Operational design questions are the low-level design round as it is asked of infrastructure engineers. Nothing here is a distributed systems question and nothing is about classes — the unit of design is a component or a workflow, and the prompts are the things an SRE, DevOps or platform engineer actually builds: a deployment pipeline, a rollout controller that can reverse itself, a health checking scheme, a secrets rotation workflow, a configuration distribution system, a backup and restore strategy. There is no capacity arithmetic and, in these questions, no code. What is being assessed is whether you have operated the thing you are describing: whether you know which failure each mechanism actually addresses, what it costs when it misfires, and what happens the first time it is used under pressure.
The recurring theme is that every one of these systems can cause the outage it exists to prevent, and interviewers listen for whether a candidate reaches that on their own. A health check that verifies its database removes the whole fleet the moment the database is slow; an automatic rollback triggered on noise flaps production between two versions; a config change propagates a typo everywhere in seconds; a backup strategy that has never been restored is a claim rather than a capability. Strong answers bound the damage a control system can do — a cap on how much capacity may be removed, a defined behaviour when its own inputs are missing, an override a human can always reach — and they distinguish the graceful path from the emergency one instead of designing a single mechanism that serves neither well. Expect a follow-up asking what happens at the moment the mechanism is needed most.
6 low-level design · 1/3/2 easy/medium/hard
No, though they overlap in vocabulary. A system design round asks how a system spans machines and is usually carried through to capacity arithmetic. An operational design round stays with one mechanism and asks how it behaves in practice: what a health check is allowed to mean, how a rollout decides to reverse itself, what happens to in-flight requests when a credential rotates. The questions are smaller in scope and considerably less forgiving about hand-waving, because the interviewer has usually been paged by the thing being described.
Rarely in this part of the loop. SRE and DevOps loops usually do include a separate coding round, but the design round described here is discussed and diagrammed rather than implemented — none of the questions on this page carry code, and the answers turn on mechanisms and trade-offs rather than on class structure.
No, deliberately. A large share of SRE and DevOps work happens on bare metal, on-premise or across providers, and a question that assumes managed services excludes the people running everything themselves. The mechanisms here — dependency graphs, credential overlap, staged propagation, log archiving — are the same whoever owns the hardware. Provider-specific service selection is covered separately under the cloud architecture questions.