# ATLAS-GATE-MCP Deployment Instructions
## TL;DR - Get Running in 5 Minutes
```bash
# 1. Make script executable
chmod +x deploy.sh
# 2. Run deployment (use your IP and SSH key path)
./deploy.sh 49.12.230.179 /home/kubuntux/Downloads/.ssh/id_rsa
# 3. Wait 2-3 minutes for everything to boot
# 4. Access at http://49.12.230.179:30000
# Health check: http://49.12.230.179:30000/health
```
Done. That's it.
---
## What the Script Does
```
./deploy.sh 49.12.230.179 /path/to/ssh/key
|
├─ [1/6] SSH into server & install k3s
| (Kubernetes in 1 binary)
|
├─ [2/6] Verify cluster is ready
|
├─ [3/6] Install nginx ingress controller
|
├─ [4/6] Create namespace + secrets
| (atlas-gate namespace)
|
├─ [5/6] Deploy all services via kubectl
| - PostgreSQL (audit logs)
| - Redis (sessions)
| - MCP server
| - persistent volumes
|
└─ [6/6] Verify everything is running
Print access info
```
---
## After Deployment
### Verify Everything Works
```bash
# SSH into server
ssh -i /path/to/key root@49.12.230.179
# Check pods
kubectl get pods -n atlas-gate
# Check services
kubectl get svc -n atlas-gate
# View logs
kubectl logs -n atlas-gate -l app=mcp-server -f
# Exit
exit
```
### Connect Your Client
#### Option 1: Port Forward (Local Testing)
```bash
# Terminal 1: SSH and forward
ssh -i /path/to/key -L 3000:localhost:3000 root@49.12.230.179
# Terminal 2: Use local connection
curl http://localhost:3000/health
```
#### Option 2: Direct Access (Production)
```bash
# Access directly from your computer
curl http://49.12.230.179:30000/health
```
#### Option 3: DNS + TLS (Advanced)
Point your domain to the server IP and configure ingress with SSL cert.
---
## File Structure
```
ATLAS-GATE-MCP/
├── deploy.sh ← Run this!
├── DEPLOY_INSTRUCTIONS.md ← This file
├── KUBERNETES_CLIENT_CONFIG.md ← Full k8s guide
├── bin/
│ └── server-network.js ← HTTP server for cloud
├── docker-compose.yml ← Local testing (optional)
└── ... other files
```
---
## Accessing Services
| Service | URL | Type |
|---------|-----|------|
| MCP API | http://49.12.230.179:30000/mcp | HTTP POST |
| Health | http://49.12.230.179:30000/health | HTTP GET |
| Metrics | http://49.12.230.179:30000/metrics | Prometheus |
| Audit | http://49.12.230.179:30000/audit/export | HTTP GET |
---
## Troubleshooting
### Script Fails at SSH
```bash
# Verify SSH key and IP
ssh -i /path/to/key root@49.12.230.179 "echo Connected"
# If it works, run deploy again
chmod +x deploy.sh
./deploy.sh 49.12.230.179 /path/to/key
```
### Pods Not Ready After 3 Minutes
```bash
ssh -i /path/to/key root@49.12.230.179
# Check what's stuck
kubectl get pods -n atlas-gate
# Describe stuck pod
kubectl describe pod <pod-name> -n atlas-gate
# Check logs
kubectl logs -n atlas-gate <pod-name>
exit
```
### Can't Connect After Deployment
```bash
# Verify service is exposed on NodePort
ssh -i /path/to/key root@49.12.230.179
kubectl get svc -n atlas-gate mcp-server-nodeport
exit
# Try direct HTTP
curl -v http://49.12.230.179:30000/health
```
### Out of Memory or Disk
```bash
ssh -i /path/to/key root@49.12.230.179
# Check resources
kubectl top nodes
kubectl top pods -n atlas-gate
# Check disk
df -h
du -sh /mnt/atlas-data/*
exit
```
---
## Environment Variables (Pre-configured)
The deployment script sets these automatically. To change them, edit the manifests in `deploy.sh`:
```
MCP_PORT=3000
MCP_BIND=0.0.0.0
MCP_ROLE=ANTIGRAVITY
DATABASE_URL=postgresql://atlas_user:atlas_password_secure_change_me@postgres:5432/atlas_gate
REDIS_URL=redis://redis:6379
WORKSPACE_ROOT=/workspace
NODE_ENV=production
AUDIT_BACKEND=postgres
SESSION_BACKEND=redis
```
**Note**: Change `atlas_password_secure_change_me` in production!
---
## Persistent Data Location
On the Hetzner server:
```
/mnt/atlas-data/
├── postgres/ (20GB - audit logs, plans, sessions)
├── redis/ (5GB - session cache)
└── workspace/ (5GB - workspace files)
```
These are mounted as Kubernetes PersistentVolumes and survive container restarts.
---
## Monitoring & Logs
### View MCP Logs
```bash
ssh -i /path/to/key root@49.12.230.179
# Follow logs in real-time
kubectl logs -n atlas-gate -l app=mcp-server -f
# View PostgreSQL logs
kubectl logs -n atlas-gate -l app=postgres -f
# View Redis logs
kubectl logs -n atlas-gate -l app=redis -f
exit
```
### Check Service Health
```bash
# From your computer
curl -v http://49.12.230.179:30000/health
# Response should be 200 OK with JSON:
# {
# "status": "healthy",
# "timestamp": "2024-02-04T...",
# "uptime": 123.45,
# "role": "ANTIGRAVITY",
# "memory": {...},
# "backends": {...}
# }
```
### Export Metrics
```bash
curl http://49.12.230.179:30000/metrics
# Prometheus-format output:
# mcp_uptime_seconds 3600
# mcp_memory_heapused_bytes 52428800
# ...
```
---
## Scaling Up
### Add More MCP Server Replicas
```bash
ssh -i /path/to/key root@49.12.230.179
# Scale to 2 replicas
kubectl scale deployment mcp-server --replicas=2 -n atlas-gate
# Verify
kubectl get pods -n atlas-gate
exit
```
### Increase Database Storage
The default is 20GB for PostgreSQL. To increase:
```bash
ssh -i /path/to/key root@49.12.230.179
# Edit the PVC
kubectl edit pvc postgres-pvc -n atlas-gate
# Change: storage: 20Gi → 50Gi
exit
```
---
## Uninstall
### Remove All Services
```bash
ssh -i /path/to/key root@49.12.230.179
# Delete the namespace (all services removed)
kubectl delete namespace atlas-gate
# Remove persistent volumes
kubectl delete pv postgres-pv redis-pv workspace-pv
# Remove data
rm -rf /mnt/atlas-data
# Optional: Uninstall k3s completely
/usr/local/bin/k3s-uninstall.sh
exit
```
---
## Next Steps
1. ✅ Run `./deploy.sh 49.12.230.179 /path/to/ssh/key`
2. Wait 2-3 minutes for pods to be ready
3. Test: `curl http://49.12.230.179:30000/health`
4. Read KUBERNETES_CLIENT_CONFIG.md for full options
5. Configure your Windsurf/Antigravity client to connect
---
## Quick Commands Reference
```bash
# Deploy
./deploy.sh 49.12.230.179 /path/to/key
# Check status
ssh root@49.12.230.179 'kubectl get pods -n atlas-gate'
# View logs
ssh root@49.12.230.179 'kubectl logs -n atlas-gate -l app=mcp-server -f'
# Test connection
curl http://49.12.230.179:30000/health
# Scale replicas
ssh root@49.12.230.179 'kubectl scale deployment mcp-server --replicas=2 -n atlas-gate'
# Uninstall
ssh root@49.12.230.179 'kubectl delete namespace atlas-gate'
```
---
Done. Run the script and you're up.
