blue/.blue/docs/spikes/0001-forgejo-tunnelless-access.md

Spike: Forgejo Access Without Tunnel

Date: 2026-01-24 Status: Complete

Problem

Currently accessing Forgejo requires kubectl port-forward. We want direct HTTPS access at git.beyondtheuniverse.superviber.com.

Current State

The infrastructure in coherence-mcp/infra is designed but not fully wired:

Component	Status	Blocker
EKS Cluster	Unknown	Need to verify deployment
Forgejo Deployment	Designed	Depends on cluster
AWS ALB Ingress	Template	${ACM_CERT_ARN} placeholder
cert-manager	Template	${ACME_EMAIL}, ${DOMAIN} placeholders
AWS LB Controller IAM	Designed	Controller not installed
DNS	Configured in PowerDNS	PowerDNS may not be deployed

Root Cause

The ingress at kubernetes/ingress/core-services.yaml uses:

alb.ingress.kubernetes.io/certificate-arn: ${ACM_CERT_ARN}

This placeholder is never substituted. Additionally, the AWS Load Balancer Controller may not be installed.

Options

Option A: ACM + AWS ALB (Current Design)

Pros: Native AWS, managed TLS, WAF integration possible Cons: Vendor lock-in, requires ACM setup, more moving parts

Steps:

1. Create ACM wildcard certificate for *.beyondtheuniverse.superviber.com
2. Install AWS Load Balancer Controller via Helm
3. Substitute ${ACM_CERT_ARN} with actual ARN
4. Apply ingress
5. Point DNS to ALB

Option B: NGINX Ingress + cert-manager + Let's Encrypt

Pros: Portable, auto-renewing certs, no ACM dependency Cons: Different from current design, requires NGINX controller

Steps:

1. Install NGINX Ingress Controller
2. Configure cert-manager with Let's Encrypt
3. Create Certificate resources for domains
4. Update ingress to use NGINX class
5. Point DNS to NGINX LoadBalancer

Option C: NLB + Pod TLS (Simplest)

Pros: Uses existing NLB, minimal changes, works today Cons: TLS at pod level, can't share certs across services

Steps:

1. Add HTTPS (443) listener to existing NLB
2. Point to Forgejo on port 3000 (or configure Forgejo for 443)
3. Use cert-manager to provision TLS cert for Forgejo
4. Mount cert in Forgejo pod
5. Configure Forgejo for TLS termination

Option D: Tailscale/Cloudflare Tunnel (Zero Infrastructure)

Pros: Works without public IP, easy setup, free tier Cons: External dependency, not self-hosted

Recommendation

Option A for production alignment with existing design. But first, verify cluster state.

Verification Steps

# 1. Check if cluster exists and accessible
aws eks describe-cluster --name alignment-production --region us-east-1

# 2. Check if kubectl works
kubectl get nodes

# 3. Check if Forgejo is deployed
kubectl get pods -n forgejo

# 4. Check if AWS LB Controller is installed
kubectl get pods -n kube-system | grep aws-load-balancer

# 5. Check if cert-manager is installed
kubectl get pods -n cert-manager

# 6. Check existing load balancers
aws elbv2 describe-load-balancers --region us-east-1

Quick Fix (If Cluster Exists)

If the cluster is running but just missing the ALB setup:

# 1. Create ACM certificate
aws acm request-certificate \
  --domain-name "*.beyondtheuniverse.superviber.com" \
  --validation-method DNS \
  --region us-east-1

# 2. Install AWS Load Balancer Controller
helm repo add eks https://aws.github.io/eks-charts
helm install aws-load-balancer-controller eks/aws-load-balancer-controller \
  -n kube-system \
  --set clusterName=alignment-production \
  --set serviceAccount.create=false \
  --set serviceAccount.name=aws-load-balancer-controller

# 3. Apply ingress with correct ARN
export ACM_CERT_ARN="arn:aws:acm:us-east-1:ACCOUNT:certificate/CERT_ID"
envsubst < kubernetes/ingress/core-services.yaml | kubectl apply -f -

# 4. Get ALB DNS name
kubectl get ingress -n ingress core-services -o jsonpath='{.status.loadBalancer.ingress[0].hostname}'

# 5. Point DNS (in PowerDNS or external DNS)
# Create CNAME: git.beyondtheuniverse.superviber.com -> ALB_DNS_NAME

Questions for User

1. Is the EKS cluster currently deployed and running?
2. Do you have Route53 managing superviber.com or is it external DNS?
3. Is PowerDNS deployed and authoritative for the subdomain?
4. Do you prefer ACM (AWS managed) or Let's Encrypt (self-managed) for TLS?

Next Steps

1. Run verification steps above
2. Choose option based on current cluster state
3. Implement chosen option
4. Update runbook to remove port-forward requirement