GKE Cluster Requirements for CI/CD

Prerequisites and setup guide for GKE clusters required by CI/CD deployment workflows

GKE Cluster Requirements for CI/CD

Overview

The CI/CD pipeline requires GKE clusters to exist before deployment workflows can succeed. This document outlines the required clusters and how to create them.

Required Clusters

Preview Cluster

Name: preview-mcp-server-langgraph-gke Location: us-central1 Project: vishnu-sandbox-20250310 Type: GKE Autopilot (recommended) or Standard

Creation via Terraform:

cd terraform/environments/gcp-preview
terraform init
terraform plan
terraform apply

Manual Creation:

gcloud container clusters create-auto preview-mcp-server-langgraph-gke \
  --region=us-central1 \
  --project=vishnu-sandbox-20250310 \
  --release-channel=regular \
  --enable-master-authorized-networks \
  --network=default \
  --subnetwork=default

Production Cluster

Name: production-mcp-server-langgraph-gke Location: us-central1 Project: vishnu-sandbox-20250310 Type: GKE Autopilot

Creation via Terraform:

cd terraform/environments/gcp-prod
terraform init
terraform plan
terraform apply

CI/CD Workflow Dependencies

deploy-preview-gke.yaml

This workflow requires:

• ✅ GKE cluster: preview-mcp-server-langgraph-gke in us-central1
• ✅ Service account: [e-mail]
• ✅ Artifact Registry: mcp-preview repository
• ✅ GitHub variable: ENABLE_PREVIEW_AUTODEPLOY=true

Workflow Trigger:

• Push to main branch (when ENABLE_PREVIEW_AUTODEPLOY=true)
• Manual dispatch

Current Status:

• ✅ Cluster created and operational

Resolution: Create the cluster using one of the methods above, then re-run the workflow.

deploy-production-gke.yaml

This workflow requires:

• ⚠️ GKE cluster: production-mcp-server-langgraph-gke in us-central1
• ✅ Service account: [e-mail]
• ✅ Artifact Registry: mcp-production repository

Workflow Trigger:

• Manual dispatch (requires approval)
• Release tags

Verification

After creating clusters, verify they’re accessible:

# Verify preview cluster
gcloud container clusters describe preview-mcp-server-langgraph-gke \
  --region=us-central1 \
  --project=vishnu-sandbox-20250310

# Verify production cluster
gcloud container clusters describe production-mcp-server-langgraph-gke \
  --region=us-central1 \
  --project=vishnu-sandbox-20250310

# Test GitHub Actions service account access
gcloud container clusters get-credentials preview-mcp-server-langgraph-gke \
  --region=us-central1 \
  --project=vishnu-sandbox-20250310

kubectl cluster-info

Terraform Module Reference

The GKE Autopilot module is located at: terraform/modules/gke-autopilot/

Features:

• Autopilot mode (Google-managed nodes)
• VPC-native networking
• Workload Identity Federation support
• Master authorized networks
• Binary authorization (optional)
• Backup agent integration
• Config Connector support

Recent Fixes:

• ✅ Duplicate lifecycle blocks merged (commit b6ac198)
• ✅ Variable validation moved to lifecycle preconditions
• ✅ Terraform validation added to pre-commit hooks

Troubleshooting

Error: Cluster not found

Symptom:

Error: NOT_FOUND: projects/vishnu-sandbox-20250310/locations/us-central1/clusters/preview-mcp-server-langgraph-gke not found

Solution: The cluster doesn’t exist. Create it using the methods above.

Error: Permission denied

Symptom:

Error: github-actions-preview does not have permission to access cluster

Solution: Verify the service account has the container.developer role:

gcloud projects get-iam-policy vishnu-sandbox-20250310 \
  --flatten="bindings[].members" \
  --filter="bindings.members:github-actions-preview*"

Error: Artifact Registry repository not found

Solution: Create the Artifact Registry repositories:

# Preview
gcloud artifacts repositories create mcp-preview \
  --repository-format=docker \
  --location=us-central1 \
  --project=vishnu-sandbox-20250310

# Production
gcloud artifacts repositories create mcp-production \
  --repository-format=docker \
  --location=us-central1 \
  --project=vishnu-sandbox-20250310

Auto-Deploy Configuration

The deployment workflows check repository variables to determine if auto-deploy is enabled:

# Enable preview auto-deploy
gh variable set ENABLE_PREVIEW_AUTODEPLOY --body "true"

# Enable dev auto-deploy
gh variable set ENABLE_DEV_AUTODEPLOY --body "true"

# Verify
gh variable list

Related Documentation

• GKE Autopilot Module
• CI/CD Overview
• GitHub Actions WIF Setup
• CI/CD Failure Analysis

Last Updated

2025-11-04 (Commit: b6ac198)

Status:

• Terraform module: ✅ Fixed and validated
• Service accounts: ✅ Created and configured
• Auto-deploy variables: ✅ Set
• GKE clusters: ✅ Created (preview-mcp-server-langgraph-gke)