================================================================================
ATLAS-GATE-MCP Cloud Deployment - Setup Summary
================================================================================
YOUR SETUP:
- Server: Hetzner cax31 (49.12.230.179)
- SSH Key: /home/kubuntux/Downloads/.ssh/id_rsa
- Cost: $3.99/month
FILES CREATED:
1. DEPLOY_INSTRUCTIONS.md (START HERE)
- Quick start guide (5 minutes to running)
- Troubleshooting
- Common commands
2. deploy.sh (MAIN SCRIPT)
- Fully automated deployment
- Installs k3s, PostgreSQL, Redis, MCP server
- Sets up persistent volumes
- Usage: ./deploy.sh 49.12.230.179 /path/to/ssh/key
3. KUBERNETES_CLIENT_CONFIG.md (REFERENCE)
- Full Kubernetes documentation
- Monitoring, scaling, backups
- Advanced configuration
- Troubleshooting details
4. bin/server-network.js
- HTTP/TCP server for cloud deployment
- Health checks, metrics, audit endpoints
- Database integration
QUICK START:
1. Make script executable:
chmod +x deploy.sh
2. Deploy to server:
./deploy.sh 49.12.230.179 /home/kubuntux/Downloads/.ssh/id_rsa
3. Wait 2-3 minutes for pods to boot
4. Test connection:
curl http://49.12.230.179:30000/health
5. Configure your client to connect to:
http://49.12.230.179:30000/mcp
WHAT GETS DEPLOYED:
✓ k3s (lightweight Kubernetes)
✓ PostgreSQL 15 (audit logs)
✓ Redis 7 (sessions)
✓ MCP Server (your app)
✓ nginx ingress controller
✓ Persistent volumes (20GB postgres, 5GB redis, 5GB workspace)
ARCHITECTURE:
Internet → Port 30000 (NodePort)
↓
nginx Ingress Controller
↓
MCP Server Pod
├→ PostgreSQL (audit)
└→ Redis (sessions)
All data persisted to /mnt/atlas-data on the server
ACCESS ENDPOINTS:
Health: http://49.12.230.179:30000/health
MCP API: http://49.12.230.179:30000/mcp
Metrics: http://49.12.230.179:30000/metrics
Audit Export: http://49.12.230.179:30000/audit/export
CONNECTING CLIENTS:
Windsurf/Antigravity should connect to:
http://49.12.230.179:30000
(See KUBERNETES_CLIENT_CONFIG.md for detailed setup)
MONITORING:
Check pod status:
ssh root@49.12.230.179
kubectl get pods -n atlas-gate
View logs:
kubectl logs -n atlas-gate -l app=mcp-server -f
Check services:
kubectl get svc -n atlas-gate
DATABASE ACCESS:
PostgreSQL:
kubectl exec -it deployment/postgres -n atlas-gate -- \
psql -U atlas_user -d atlas_gate
Redis:
kubectl exec -it deployment/redis -n atlas-gate -- \
redis-cli
SCALING:
Add more servers (replicas):
kubectl scale deployment mcp-server --replicas=2 -n atlas-gate
Increase database storage:
kubectl edit pvc postgres-pvc -n atlas-gate
(Change: storage: 20Gi → 50Gi)
TROUBLESHOOTING:
If deployment fails:
1. Verify SSH works: ssh -i key root@49.12.230.179
2. Check k3s installed: ssh root@49.12.230.179 'k3s --version'
3. View script output for errors
4. Run deploy.sh again
If pods don't start:
1. Check status: kubectl get pods -n atlas-gate
2. Describe stuck pod: kubectl describe pod <name> -n atlas-gate
3. View logs: kubectl logs -n atlas-gate <pod-name>
4. Common issue: waiting for database/redis to be ready (takes ~30s)
If can't connect:
1. Verify service is exposed: kubectl get svc -n atlas-gate
2. Check NodePort: should be 30000
3. Test: curl http://49.12.230.179:30000/health
4. Check firewall: port 30000 must be accessible
PERSISTENCE:
All data is stored on the server:
/mnt/atlas-data/postgres (20GB)
/mnt/atlas-data/redis (5GB)
/mnt/atlas-data/workspace (5GB)
Data survives pod restarts, server reboots, etc.
BACKUPS:
PostgreSQL backup:
kubectl exec deployment/postgres -n atlas-gate -- \
pg_dump -U atlas_user atlas_gate | gzip > backup.sql.gz
Redis backup:
kubectl exec deployment/redis -n atlas-gate -- \
redis-cli BGSAVE
Copy from server:
scp -i key root@49.12.230.179:/mnt/atlas-data/postgres/* ./backups/
SECURITY NOTES:
⚠ Default password in deploy.sh: "atlas_password_secure_change_me"
Change it in production:
1. Edit deploy.sh (line with POSTGRES_PASSWORD)
2. Update secret: kubectl create secret... -n atlas-gate
3. Restart pods
⚠ Exposed on public IP
Consider:
- SSH tunnel for access
- Firewall rules
- TLS/HTTPS setup
- Rate limiting (already configured)
NEXT STEPS:
1. Read DEPLOY_INSTRUCTIONS.md (5-minute guide)
2. Run ./deploy.sh
3. Test with curl
4. Configure your clients
5. Monitor with kubectl logs -f
6. Refer to KUBERNETES_CLIENT_CONFIG.md for advanced setup
================================================================================
Questions? Check the documentation files:
- DEPLOY_INSTRUCTIONS.md (quick start)
- KUBERNETES_CLIENT_CONFIG.md (reference)
- CLOUD_DEPLOYMENT_GUIDE.md (architecture details)
