W3DevOps Module · Google Cloud: Resource Hierarchy, Compute Engine, Cloud Storage and GKE

Google Cloud is where my projects live: Nora Labs and CryptoEdge on Cloud Run, Quintrix ITSO on a Compute Engine VM, and my Coolify control plane on an e2 instance in us-central1-b. This is the Google Cloud module of the course — 20 lab captures and the gcloud commands I keep coming back to.

What this module covers

• The four-tier resource hierarchy and why billing budgets and alerts come first
• Cloud Shell and essential gcloud commands
• Creating an e2-medium Ubuntu VM from the CLI, and Managed Instance Groups with autoscaling
• The four GCS storage classes and lifecycle rules
• GKE Autopilot vs Standard

Lessons in this module

1. Google Cloud Platform (GCP) Fundamentals — Google Cloud Platform provides enterprise-grade global infrastructure built on the exact same high-speed networks and data centers that power Google Search, YouTube, and Gmail.
2. Google Compute Engine (GCE) & Instance Groups — Compute Engine delivers high-performance virtual machines running on Google's Andromeda SDN network. With custom machine shapes, you can pick any combination of vCPU cores and RAM.
3. Google Cloud Storage (GCS) — Google Cloud Storage (GCS) is an exabyte-scale object storage service providing global edge caching, strong consistency out of the box, and automated lifecycle policies.
4. Google Kubernetes Engine (GKE) & Operations — Google invented Kubernetes. GKE is the premier managed Kubernetes implementation in the industry, featuring four-way autoscaling (horizontal pod autoscaling, vertical pod autoscaling, node…

Lessons from deploying real projects on GCP

• Set a budget alert before anything else. It's the first lab in this track for a reason.
• New projects have no Cloud Build default service account. gcloud builds submit fails with Unknown service account on a fresh project. Nora Labs' deploy script therefore builds locally for linux/amd64 and pushes straight to Artifact Registry until a builder SA with roles/cloudbuild.builds.builder exists.
• Model availability is per region. gemini-2.5-flash resolves in us-central1; a model 404 elsewhere is caught and Nora's deterministic engine answers instead.
• Mount a bucket for state. Cloud Run's filesystem is ephemeral — Nora mounts a GCS bucket at /data so saved trips survive instance recycling.

Next in the series: from bare metal to Nginx, TLS and CDNs.

---

Keep going: Start the module → · Full course outline · Self Guide