Opscraft
Senior DevOps interviews · Lesson 3 of 10 · 5 min read

Answer frameworks

STAR for behavioural and incident questions

  • Situation — context and stakes, briefly.
  • Task — what you were responsible for.
  • Action — what you did, step by step, with technical detail.
  • Result — outcome with numbers, and what changed afterwards.
Enlarge diagramSituationcontextTaskyour jobActionwhat YOU did, step by stepResultnumbers + lessonabout 2 minutes in total
Figure 1. STAR for behavioural and incident stories. Spend most of the time on Action, and finish with the lasting improvement.
Swipe sideways to see all of it, or tap to open it full screen.

Keep it to about two minutes. End with the lesson or the permanent improvement — that's what makes it senior.

Technical "explain X" questions

  1. One-sentence definition.
  2. How it works (the mechanism).
  3. When you'd use it / trade-offs.
  4. A real example from your work, or a gotcha.
Enlarge diagramDefineone sentenceMechanismhow it worksWhen / trade-offsvs the alternativesExamplefrom your work, or agotcha
Figure 2. Answering "explain X": four short moves.
Swipe sideways to see all of it, or tap to open it full screen.

Example: "What's IRSA?" → definition → OIDC/STS flow → least privilege, vs node roles and Pod Identity → "we used it for External DNS and Flux's ECR access".

Troubleshooting questions: narrate a method

Interviewers care about how you think. Say your steps aloud:

Enlarge diagramClarifywho, since when,what changedMitigateroll back, fail overNarrow downlayer by layer,evidenceRoot causethen fixPreventalerts, tests,runbooks
Figure 3. Troubleshooting out loud: interviewers grade the method more than the final answer.
Swipe sideways to see all of it, or tap to open it full screen.
  1. Clarify scope and impact: who's affected, since when, what changed.
  2. Mitigate first if users are hurting: roll back, fail over, scale.
  3. Narrow down layer by layer with evidence (commands, metrics, logs).
  4. Root cause, then fix.
  5. Prevent: alerts, tests, guardrails, runbooks.

"First I'd check what changed — deploys in the last hour. If a deploy correlates and errors persist, I'd roll back via a Git revert. Then I'd look at ELB vs target 5xx..."

System design questions

  1. Clarify requirements — scale, SLOs, compliance, team size, existing tools, budget. Ask 2–4 good questions; state assumptions.
  2. High-level design — draw the main components and flows.
  3. Deep dive where it matters — security, failure modes, scaling, data.
  4. Trade-offs — what you chose and what you rejected, and why.
  5. Operations — monitoring, alerting, runbooks, cost, how teams use it.
Enlarge diagramClarifyscale, SLOs,compliance,budgetHigh levelcomponents andflowsDeep divesecurity, failure,scalingTrade-offschosen vsrejectedOperationsmonitoring, cost,runbooks12345
Figure 4. A system design round in five steps. There is no single right architecture; reasoning and trade-offs are what get graded.
Swipe sideways to see all of it, or tap to open it full screen.

Whiteboard rounds grade: requirement clarification, sound trade-offs, reliability thinking and communication — not a single "correct" architecture.

When you don't know

Say what you do know, how you'd find out, and reason aloud: "I haven't used Istio's ambient mode in production. My understanding is it moves the proxy out of sidecars into per-node components... To decide, I'd check..." Never bluff — senior interviewers probe until bluffing collapses, and honesty scores well.

If you realise an earlier answer was wrong, go back and correct it briefly — it shows self-awareness.