Interview questions in DevOps & Cloud, page 16
High-throughput serverless stream processing
Partition by key for per-user ordering, use a sharded log with batched consumers for backpressure, and tune batch size and shards for cost.
How would you automate canary deployment and what metrics decide rollback?
This tests progressive delivery maturity. Strong answers cover traffic splitting (10% to 100%), automated 5-10 minute health gates, and rollback triggers like error rate and p99 latency.
Systematically reduce noisy alert toil
Inventory alerts, measure frequency, actionability, and time cost, then prioritize by volume times effort.
What do PersistentVolume accessModes mean?
RWO mounts read-write by one node, ROX read-only by many nodes, RWX read-write by many nodes; block storage usually only supports RWO while shared filesystems enable RWX.
What is Infrastructure as Code?
IaC defines infra in version-controlled files, giving repeatability, peer review, and drift-free consistency, unlike error-prone console clicks.
How can a service mesh facilitate canary or A/B testing?
Tests mesh-level traffic control decoupled from app releases. Strong answers name traffic shifting, request routing, ingress gateways, and telemetry-driven rollback. Red flag: citing mTLS alone and omitting observability automation.
Design automated microservice provisioning workflow
Template scaffolding plus a pipeline that creates repo, CI/CD, and infra as code, with idempotent steps and rollback.
How does a StatefulSet give stable identity and storage?
Ordinal Pod names plus a headless Service yield stable per-Pod DNS; volumeClaimTemplates give each ordinal its own persistent PVC that follows it on reschedule.
Automate patching across a VM fleet
Use a patch or config tool to target by tag, roll out in canaried waves with health checks, and prefer immutable golden images long term.
Canary vs shadow deployments: use cases and requirements
This tests whether you distinguish user-facing rollouts from invisible duplication. Canary routes some real users to new code to limit blast radius; shadow mirrors traffic to an isolated clone to test performance without user impact.
Design a centralized auto-remediation platform
Event ingestion, a rules engine mapping alerts to playbooks, a sandboxed execution runtime, and guardrails like dry-run, rate limits, and rollback.
Why is a Pod with a PVC stuck Pending?
PVC may be unbound from missing StorageClass, no matching PV, mismatched access mode or size, zone or capacity limits, or WaitForFirstConsumer; diagnose with describe on Pod and PVC plus events.
Centralized logging across microservices
Ship structured logs from every service into a central searchable store, then propagate a correlation ID through all hops to trace one request.
Blue/green deployment fails during switch-over with partial decommissioning; recovery and process changes?
Tests whether you can recover when a blue/green rollback path is compromised. Strong answers stop the bleed, revive blue if possible, and mandate keeping blue fully warm until green is stable. Red flag: "just roll back" ignoring partial decommissioning.
Measure ROI of toil reduction efforts
Track toil hours, percent of time on toil, incidents auto-resolved, and engineer cost saved, then frame as ROI and risk reduction.
How does a StatefulSet recover a Pod after node failure?
Node goes NotReady, Pod is marked for deletion, the same-ordinal Pod is recreated and reattaches its existing PVC from volumeClaimTemplates, preserving data; safety needs the old Pod confirmed…
State drift in Terraform
Drift is when live infrastructure diverges from recorded state, usually via manual console changes; detect with plan or refresh, remediate by re-applying or importing.
How do you deploy a hotfix during a multi-stage canary release?
Dark-launch the fix to the canary cohort, preserve metrics, then jointly promote.
Blue-green deploys with schema migrations
The shared database means both versions hit one schema, so breaking changes must be split into backward-compatible steps via expand-and-contract.
How do you resize a live PersistentVolume?
Edit the PVC's requested size upward; the StorageClass must set allowVolumeExpansion true and the CSI driver must support expansion, ideally online so no Pod restart is needed; shrinking is not allowed.
We are hiring for this. Every open role lists the topics its interview covers, so you can prepare for the real thing rather than guessing.
See open roles