Cloud / Domain Knowledge Oral Round (GCP / AWS / K8s)

Salesforce GCP / Public-Cloud-team loops include a dedicated domain-knowledge round drilling on cloud platform internals and Kubernetes — even if the candidate's background is on a different cloud. Concept fluency on K8s, cloud pipelines, and (where relevant) cloud-migration experience. Requirements Verbal-only round; no coding (but interviewers may sketch on the shared whiteboard). Topics drilled vary by team: Kubernetes fundamentals: pods vs deployments vs services, scheduling, autoscaling (HPA / VPA / Cluster Autoscaler), networking (CNI, services, ingress), storage (PV / PVC / StorageClass), troubleshooting (kubectl describe, events, common pod failure modes). Cloud platform: GCP-equivalents are weighted on GCP team loops; AWS-equivalents on candidates with AWS background. EC2 / GCE, S3 / GCS, IAM, VPC, security groups / firewall rules, load balancers, managed databases. Operational detail ("how do you actually set up Y") not just concepts. CI/CD and pipelines: deployment pipelines, blue-green vs canary, infra-as-code (Terraform, Pulumi), GitOps (ArgoCD, Flux). Cloud migration: specific to candidates who have done one — what got migrated, how data was moved, what broke, how it was validated. Domain-specific architectures: e.g. pipeline system design for the GCP organisation. Round structure: rapid-fire concept questions for 15-20 minutes, then one or two open-ended scenario / architecture questions ("design a deployment pipeline for X", "diagnose this incident"). Notes K8s answers interviewers reward: Pod lifecycle (Pending → ContainerCreating → Running → Succeeded/Failed/CrashLoopBackoff). Common failure modes per state. Difference between Deployment, StatefulSet, DaemonSet — and when each is right. Service types (ClusterIP, NodePort, LoadBalancer, ExternalName, headless service). Scheduling: requests vs limits, taints / tolerations, node affinity, pod-affinity / anti-affinity. Autoscaling: HPA (per-pod horizontal), VPA (per-pod vertical), CA (node count). Know which signal each consumes. Cloud answers: know the difference between region / zone / AZ, object storage consistency (S3 strong read-after-write since 2020), managed-DB failover semantics, IAM role assumption flow. Salesforce GCP team interviewers explicitly probe on whichever cloud the candidate's resume lists — don't bluff on a cloud you haven't operated. Cloud migration: be specific about what data, what cutover strategy (dual-write, shadow read, hard cutover), what monitoring you added during cutover, what rolled back. "We did a cloud migration" with no specifics is the failure mode here. Pipeline system design: same expectations as a regular SD round (ingestion → processing → serving), but the interviewer expects you to name concrete cloud services. Cross-cloud probing happens: an AWS-experienced candidate may be asked AWS questions and judged on whether they actually know AWS deeply, not on whether they know GCP. Honesty about cloud experience scores better than improvisation. Preparation Refresh K8s fundamentals: CKAD-style flashcards (objects, controllers, networking, storage, troubleshooting). For each cloud you list on your resume, be able to walk through one full architecture you built end-to-end — services used, scale, what you'd do differently. If migration experience is on your resume, prep a 5-minute migration story with timeline, blockers, rollback plan, validation. Refresh one IaC tool (Terraform) at the level of: how state works, how plan/apply differs, how to handle drift, how to manage multi-environment workspaces. For the pipeline SD sub-prompt, have a Kafka / Pub-Sub → Flink / Dataflow → BigQuery / Snowflake reference architecture ready.