The skills a DevOps interview actually tests

These are preparation categories, not a published scoring rubric. Use the job description and recruiter briefing to decide how much time each deserves. The prompts below are editorial exercises.

Start with the user impact and one testable hypothesis. Name the observation that would change your mind, collect it and choose the next check. During an incident, a reversible mitigation may need to precede root-cause analysis.

Practise a CrashLoopBackOff, a service returning 503 and a high p99 with a low average. Explain why each observation narrows the investigation.

Trace behavior through the abstraction: a container still uses kernel process and memory mechanisms. Practise signals, cgroup limits, TCP state and DNS resolution. Verify the cause of exit 137 rather than assuming every SIGKILL was an OOM kill.

For a time-based availability SLO over exactly 30 days, the window contains 43,200 minutes. A 99.9% objective permits 43.2 minutes of bad time; 99.95% permits 21.6. A request-based SLO instead budgets bad requests, so outage minutes cannot be inferred without traffic assumptions.

Define the measured event, success condition and window. Then discuss an error-budget policy. Google's SLO workbook provides the operational context; the arithmetic here is a worked example.

State units before sizing. Moving five decimal petabytes over one dedicated 10Gbps link takes 4,000,000 seconds, about 46.3 days at line rate. That excludes protocol overhead and storage limits. Use the result to challenge a four-hour full-restore proposal.

Explain how to limit the affected users and how to detect a bad change. Test a rollback, including the case where a schema migration makes binary rollback unsafe. Practise describing the checks before promoting an immutable artifact.

Write a small utility with explicit error handling, bounded input handling and a safe retry policy. Test the second run as well as the first. Ask whether partial completion should resume, roll back or require intervention.

Inspect the bill before choosing a saving. Identify usage, unit rate and allocation, then explain a change that preserves the service objective. Distinguish a discount commitment from a capacity reservation.

Turn a technical failure into an operational consequence, estimate the uncertainty and make a recommendation. Practise explaining a migration to someone who must approve its cost but does not maintain the system.