Behavioural questions with model answers
"Tell me about a production incident you handled."
Structure: impact and detection → diagnosis with real commands/evidence → mitigation → root cause → prevention. Example shape:
"[Situation] After a release, checkout latency spiked and 5xx rose to about [X]% during peak. [Task] I was on call. [Action] I checked what changed — a deploy 20 minutes earlier. Pods were restarting;
describeshowed OOMKilled, exit 137; memory graphs climbed steadily from the deploy. I rolled back by reverting the image tag commit and Flux reconciled within a couple of minutes; errors dropped. Then we found a cache without eviction in the new code. [Result] Impact lasted [N] minutes. Afterwards we added memory alerts at 90% of the limit, a load test step in CI, and canary analysis including memory."
"Tell me about a disagreement with a senior or a developer."
"A senior engineer wanted to give every team cluster-admin to speed things up. I was worried about blast radius. I gathered data — which actions teams actually needed — and proposed namespace-scoped roles plus a self-service path for the few cluster-level needs. We discussed trade-offs openly; we agreed on the scoped approach with a two-week review. Once decided, I committed fully, including the parts I'd argued against."
Key ideas: listen first, use data, focus on the goal, disagree and commit.
"Describe a failure."
Own it, explain what you learned and what you changed: e.g. a config change applied without a plan review that caused downtime → you introduced mandatory plan review in PRs and policy checks.
"How do you prioritise when everything is urgent?"
"By impact and urgency: active incidents and anything blocking production come first. Then I make trade-offs explicit — tell stakeholders what will slip and why, agree priorities with my manager, and avoid silently juggling. I protect time for reliability work so we're not always firefighting."
"Tell me about a time you reduced cost."
Baseline → analysis → change → savings with numbers → trade-offs: "EKS spend was [₹/$X]/month; Cost Explorer and Kubecost showed requests at ~3× actual usage and all on-demand nodes. We right-sized requests using VPA recommendations, introduced Karpenter with Spot for stateless workloads and Graviton nodes. Spend dropped ~[Y]% with no SLO impact; we kept stateful and system pods on-demand."
"Why are you leaving?" / "Why this company?"
Positive and forward-looking — scope, scale, learning — never criticise your employer:
"I've learned a lot at [current company], especially around [e.g. EKS and GitOps]. I'm looking for larger scale and deeper platform ownership — more services, more traffic, and building internal developer platforms — and your team's [specific thing from their JD/blog] is exactly that. [If you're relocating: the new city also makes sense for me personally.]"
"How do you stay current?"
Specific sources and hands-on proof: Kubernetes/EKS release notes and blogs (AWS Containers blog, CNCF), KubeCon talks, newsletters, and projects — "I built the Terraform EKS repo and a PR-validation GitHub Action to try things properly", plus CKA and AWS SA certifications.
"What does a good on-call look like to you?"
Actionable alerts tied to SLOs (not CPU noise), runbooks for every alert, fair rotations, blameless postmortems with owned action items, and steadily reducing toil so on-call gets quieter over time.
"Why should we hire you for Senior DevOps?"
Connect your strengths to their needs: "You run EKS at scale with GitOps — that's what I do daily. I bring hands-on Kubernetes depth (CKA), AWS architecture (SA), CI/CD across Jenkins and GitHub Actions, and a habit of measuring impact. Where I'm still building depth, like Terraform at scale, I've shown I can ramp up quickly through real projects."