Google Cloud (GCP) interview questions with answer outlines

Fifteen common Google Cloud interview questions

1. Explain the GCP resource hierarchy: organisation, folders, projects, resources.
2. How does IAM work in GCP? Basic, predefined and custom roles?
3. What are service accounts, and how should workloads authenticate?
4. Compute Engine versus App Engine versus Cloud Run versus GKE?
5. Cloud Storage classes and when to use each?
6. Cloud SQL versus Spanner versus Firestore versus Bigtable?
7. How does BigQuery work, and how do you control its costs?
8. What is Pub/Sub, and what delivery guarantees does it offer?
9. How is a GCP VPC different from other clouds' networks?
10. What does a global load balancer give you?
11. How do you secure data with VPC Service Controls?
12. How do you monitor services on GCP?
13. How do you deploy infrastructure as code on GCP?
14. How do you reduce GCP costs?
15. Design a data pipeline from events to dashboards on GCP.

Hierarchy and IAM

Policies set at the organisation or folder level are inherited by projects and resources, so structure folders around environments or business units and apply guardrails high up with organisation policies. Grant predefined roles rather than broad basic roles (Owner, Editor, Viewer), and use custom roles when needed. Workloads should use service accounts with least privilege, ideally through Workload Identity on GKE or attached service accounts, rather than downloaded keys.

Compute and data choices

BigQuery and Pub/Sub

• BigQuery separates storage and compute and bills on-demand by data scanned or through reserved capacity. Control costs by partitioning and clustering tables, selecting only needed columns, using previews instead of SELECT *, and setting quotas.
• Pub/Sub is a global messaging service with at-least-once delivery by default (an exactly-once option exists for pull subscriptions in a region), so consumers should be idempotent. It decouples producers from consumers and feeds Dataflow or Cloud Run.

Networking

GCP VPCs are global, with regional subnets, so one network can span regions without peering. Shared VPC lets a host project provide networking to service projects centrally. The global external HTTP(S) load balancer offers a single anycast IP that routes users to the nearest healthy backend worldwide. VPC Service Controls create perimeters around services such as Cloud Storage and BigQuery to reduce data exfiltration risk, complementing IAM.

Operations and cost

Cloud Monitoring and Cloud Logging (the Cloud Operations suite) cover metrics, logs, traces and alerting. Use Terraform or Infrastructure Manager for infrastructure as code. Reduce cost with committed-use and sustained-use discounts, spot VMs, Cloud Run's scale-to-zero, storage lifecycle rules and BigQuery cost controls. A typical event pipeline: Pub/Sub ingests events, Dataflow transforms them, BigQuery stores them, and a BI tool such as Looker serves dashboards. See data warehouse interview questions.

Certifications and interviews

Google Cloud certifications cover the service catalogue well and help with screening, but interviews go beyond them: they ask why you chose a service, what went wrong in production and how much it cost. Pair certification study with one hands-on project — a containerised service on Cloud Run with a Cloud SQL database and BigQuery reporting is a good scope — so you have real decisions and trade-offs to discuss.

Practising

Answer these out loud, in two minutes or less each, and expect a follow-up on every claim. A free AI interview assistant such as CoPilot Interview can run Google Cloud questions as a mock interview built from your résumé and score each answer. Prepare the event-to-dashboard pipeline as a design you can explain in five minutes. See also Kubernetes interview questions.

Disclosure: CoPilot Interview is our product, so weigh what we say about it accordingly. Everything else on this page applies whichever tool you use, or none.

Frequently asked questions

What are common GCP interview questions?

Resource hierarchy, IAM and service accounts, compute options, storage classes, Cloud SQL versus Spanner versus Firestore versus Bigtable, BigQuery, Pub/Sub, global VPCs and load balancing, VPC Service Controls, monitoring and cost.

Cloud Run or GKE?

Cloud Run suits stateless containers that should scale automatically, including to zero, with minimal operations. GKE suits complex platforms needing full Kubernetes features and control.

How do you reduce BigQuery costs?

Partition and cluster tables, query only necessary columns, avoid SELECT * on large tables, use table previews, set quotas, and consider reserved capacity for steady heavy usage.

What makes GCP networking different?

VPC networks are global with regional subnets, so a single network can span regions, and the global load balancer provides one anycast IP for worldwide traffic.