Google Cloud Platform (GCP) Skill

Service Overview

Use the dedicated terraform skill when the question is primarily about Terraform layout, state boundaries, brownfield import/export, workspaces, or CI-driven plan/apply workflows. Keep this gcp skill focused on Google Cloud services and gcloud-centric operations.

Core Services

• Compute:
  • Cloud Run: Serverless containers (default choice for stateless apps).
  • GKE: Managed Kubernetes for complex orchestrations.
  • Compute Engine: Raw VMs for specific OS/kernel needs.
• Data & Storage:
  • Cloud Storage (GCS): Object storage.
  • Cloud SQL: Managed PostgreSQL/MySQL/SQL Server.
  • BigQuery: Serverless data warehouse (analytics).
  • Firestore: NoSQL document database.
• AI/ML:
  • Vertex AI: Unified platform for models (Gemini, PaLM), training, and deployment.

<workflow>

gcloud CLI & Scripting

Configuration & Auth

<guardrails>

Avoid interactive prompts in scripts.

</guardrails> <example>

# Production/CI: Use Service Account Key or Workload Identity
gcloud auth activate-service-account --key-file=key.json

# Local Dev: User Login
gcloud auth login
gcloud config set project MY_PROJECT_ID

</example>

Scripting Best Practices

1. Structured Output

Never parse default text output. Use --format (json/yaml) and --filter.

<example>

# Bad
gcloud compute instances list | grep RUNNING

# Good (Parseable JSON)
gcloud compute instances list --format="json"

# Good (Filter + Specific Value)
gcloud run services list \
  --filter="status.conditions.status=True AND metadata.name:my-service" \
  --format="value(status.url)"

</example>

2. Deterministic Filters

Flatten complex resources to find what you need.

<example>

# Find latest revision of a service
gcloud run revisions list \
  --service=my-service \
  --sort-by="~metadata.creationTimestamp" \
  --limit=1 \
  --format="value(metadata.name)"

</example>

3. Quiet Mode

Suppress "updates available" warnings and prompts.

<example>

export CLOUDSDK_CORE_DISABLE_PROMPTS=1
gcloud ... --quiet

</example>

Automation Patterns

1. Cloud Run Deployment

Standard pattern for deploying containers.

<example>

gcloud run deploy my-service \
  --image gcr.io/my-project/my-image:tag \
  --platform managed \
  --region us-central1 \
  --allow-unauthenticated \
  --set-env-vars="DEBUG=true,DB_HOST=10.0.0.2"

</example>

2. Secret Management

<guardrails>

Access secrets securely (requires Secret Manager API).

</guardrails> <example>

# Mount as volume in Cloud Run (Preferred)
gcloud run deploy ... --set-secrets="/secrets/db=my-db-secret:latest"

# Access via CLI (for ops scripts)
gcloud secrets versions access latest --secret="my-secret"

</example> </workflow>

AlloyDB

AlloyDB is a fully managed PostgreSQL-compatible database with columnar engine and ML-assisted auto-vacuum.

Cluster / Instance Model

• Cluster: regional resource containing one primary and optional read pool instances.
• Primary instance: read-write; choose machine type and vCPUs.
• Read pool: horizontally scalable read-only replicas within the same cluster.

# Create a cluster
gcloud alloydb clusters create my-cluster \
  --region=us-central1 \
  --password=SECRET \
  --network=projects/MY_PROJECT/global/networks/default

# Create primary instance
gcloud alloydb instances create my-primary \
  --cluster=my-cluster \
  --region=us-central1 \
  --instance-type=PRIMARY \
  --cpu-count=4

PSA Networking Requirement

AlloyDB requires Private Service Access (PSA) — a peered VPC range allocated for Google-managed services. Client VMs must be in the same VPC (or a connected VPC) to reach the instance IP.

# Allocate PSA range (one-time per VPC)
gcloud compute addresses create google-managed-services-default \
  --global \
  --purpose=VPC_PEERING \
  --prefix-length=20 \
  --network=default

# Create the peering
gcloud services vpc-peerings connect \
  --service=servicenetworking.googleapis.com \
  --ranges=google-managed-services-default \
  --network=default

AlloyDB vs Cloud SQL

Aspect	AlloyDB	Cloud SQL
Engine	PostgreSQL-compatible only	PostgreSQL, MySQL, SQL Server
Performance	~4× higher throughput (columnar engine, shared memory cache)	Standard managed RDBMS
HA	Auto-failover < 60 s, cross-zone	Regional replica, ~60 s failover
Pricing	Higher; compute + storage separate	Instance + storage (simpler)
Best for	High-throughput OLTP, mixed OLTP/OLAP	General-purpose managed SQL

---

Secret Manager Patterns

Diff-Based Updates

Avoid creating unnecessary secret versions. Compare the current value before adding a new version.

# Read existing value
CURRENT=$(gcloud secrets versions access latest --secret="my-secret" 2>/dev/null || echo "")

NEW_VALUE="new-password-here"

if [ "$CURRENT" != "$NEW_VALUE" ]; then
  echo -n "$NEW_VALUE" | gcloud secrets versions add my-secret --data-file=-
  echo "Secret updated."
else
  echo "Secret unchanged, skipping version creation."
fi

Common Access Patterns

# Access the latest version
gcloud secrets versions access latest --secret="my-secret"

# Access a specific version
gcloud secrets versions access 3 --secret="my-secret"

# List versions
gcloud secrets versions list my-secret

# Create a new secret
echo -n "my-value" | gcloud secrets create my-secret \
  --data-file=- \
  --replication-policy=automatic

---

IAM Workload Identity

Workload Identity lets GKE or Cloud Run workloads impersonate a GCP service account without key files.

Annotation + Binding Chain

# 1. Create a GCP Service Account (GSA)
gcloud iam service-accounts create my-app-sa \
  --display-name="My App SA"

# 2. Grant required roles to the GSA
gcloud projects add-iam-policy-binding MY_PROJECT \
  --member="serviceAccount:my-app-sa@MY_PROJECT.iam.gserviceaccount.com" \
  --role="roles/secretmanager.secretAccessor"

# 3. Allow the Kubernetes Service Account (KSA) to impersonate the GSA
gcloud iam service-accounts add-iam-policy-binding \
  my-app-sa@MY_PROJECT.iam.gserviceaccount.com \
  --role="roles/iam.workloadIdentityUser" \
  --member="serviceAccount:MY_PROJECT.svc.id.goog[NAMESPACE/KSA_NAME]"

# 4. Annotate the KSA
kubectl annotate serviceaccount KSA_NAME \
  --namespace=NAMESPACE \
  iam.gke.io/gcp-service-account=my-app-sa@MY_PROJECT.iam.gserviceaccount.com

Service Account Impersonation (CLI)

# Impersonate a GSA from a user or another SA
gcloud storage ls \
  --impersonate-service-account=my-app-sa@MY_PROJECT.iam.gserviceaccount.com

# Generate a short-lived token
gcloud auth print-access-token \
  --impersonate-service-account=my-app-sa@MY_PROJECT.iam.gserviceaccount.com

---

VPC Networking

PSA Ranges

See AlloyDB section above. PSA is also required for Cloud SQL private IP and Memorystore.

Cloud NAT / Router

Cloud NAT allows VMs without external IPs to reach the internet.

# Create a Cloud Router
gcloud compute routers create my-router \
  --region=us-central1 \
  --network=default

# Attach Cloud NAT
gcloud compute routers nats create my-nat \
  --router=my-router \
  --region=us-central1 \
  --auto-allocate-nat-external-ips \
  --nat-all-subnet-ip-ranges

Firewall Rules

# IAP TCP tunneling (SSH/RDP via IAP)
gcloud compute firewall-rules create allow-iap-ssh \
  --network=default \
  --allow=tcp:22 \
  --source-ranges=35.235.240.0/20 \
  --description="Allow SSH via IAP"

# GCP load balancer health checks
gcloud compute firewall-rules create allow-health-checks \
  --network=default \
  --allow=tcp \
  --source-ranges=130.211.0.0/22,35.191.0.0/16 \
  --description="Allow GCP health check probers"

Purpose	CIDR
IAP TCP forwarding	35.235.240.0/20
GCP health check probers	130.211.0.0/22, 35.191.0.0/16

---

Cloud Batch

Cloud Batch is a fully managed service for batch and HPC workloads. It provisions, schedules, and autoscales VMs (including Spot/preemptible) without managing a cluster.

When to use Cloud Batch vs GKE:

Aspect	Cloud Batch	GKE
Workload type	Batch jobs, array jobs, MPI	Long-running services, microservices
Cluster management	None (fully managed)	Cluster lifecycle managed by operator
Spot/preemptible	Built-in, first-class	Node pool configuration
GPU / HPC support	A100/H100, HPC VM families	Any accelerator, custom node pools
Scheduling	Queue-based, job arrays	Kubernetes scheduler

# Submit a simple batch job from JSON spec
gcloud batch jobs submit my-job \
  --location=us-central1 \
  --config=job.json

---

References Index

• IAM Guide - Service accounts, role bindings, Workload Identity, and IAM best practices.

Cross-References

• Gemini CLI Extensions org: https://github.com/gemini-cli-extensions — community-built Gemini CLI extensions for GCP services and tooling.

Documentation & References

• SDK Cheat Sheet: gcloud cheat-sheet
• Core Services List: Google Cloud Products
• CLI Reference: gcloud CLI docs

Official References

• https://docs.cloud.google.com/
• https://docs.cloud.google.com/sdk/gcloud/reference
• https://cloud.google.com/sdk/gcloud/reference/run/deploy
• https://docs.cloud.google.com/sdk/docs/authorizing
• https://cloud.google.com/artifact-registry/docs/transition/transition-from-gcr
• https://cloud.google.com/vertex-ai/generative-ai/docs/migrate/migrate-palm-to-gemini

Shared Styleguide Baseline

• Use shared styleguides for generic language/framework rules to reduce duplication in this skill.
• General Principles
• GCP Scripting
• Bash
• Keep this skill focused on tool-specific workflows, edge cases, and integration details.

<validation> ## Validation

Add validation instructions here. </validation>