LearnHouse Deployment Troubleshooting Guide

Current Status: Port Configuration Fixed ✅

Recent Changes Made:

Fixed Port Mismatch Issue - The root cause of "no available server":
- Changed Docker Compose from port 3000 → 80
- Updated healthcheck from port 3000 → 80
- Added explicit Traefik port label: traefik.http.services.*.loadbalancer.server.port=80
Enhanced Start Script (extra/start.sh):
- Added explicit port assignments: PORT=8000, LEARNHOUSE_PORT=9000
- Fixed backend startup: Uses uvicorn app:app --host 0.0.0.0 --port 9000
- Fixed frontend startup: Uses Next.js standalone server on port 8000
Added Debug Capabilities:
- Created debug-services.sh script for troubleshooting
- Script checks PM2 processes, port usage, service connectivity

Current Architecture:

Internet → Coolify/Traefik → Container:80 → Nginx → {
  ├── Frontend (Next.js standalone): localhost:8000
  └── Backend API (FastAPI): localhost:9000
}

Network Isolation Configuration:

DEV deployment: DEPLOYMENT_NAME=dev → dev-network
LIVE deployment: DEPLOYMENT_NAME=live → live-network
Each deployment has isolated databases, Redis instances, and networks

Environment Variables Required:

See COOLIFY_ENV_VARS.md for complete list. Key variables for isolation:

DEPLOYMENT_NAME=live (or dev)
LEARNHOUSE_COOKIE_DOMAIN=edu.adradviser.ro
LEARNHOUSE_SQL_CONNECTION_STRING (separate for each deployment)
LEARNHOUSE_REDIS_CONNECTION_STRING (separate for each deployment)

Current Error Status:

✅ Port mismatch fixed: Changed from 3000 to 80
✅ Container accessibility: Traefik can now route to port 80
✅ Frontend running: Next.js server operational on port 8000
✅ Backend running: FastAPI server operational on port 9000
❌ Cross-deployment contamination: LIVE calling DEV APIs and vice versa
⚠️ Root cause: Frontend build-time API URLs not properly isolated

Identified Issues & Fixes:

Problem: Cross-deployment data contamination (LIVE sees DEV data) Root Cause: Next.js build embeds API URLs at build-time, both deployments may share same URLs Solution: Added runtime API URL patching in Docker container startup

Current Fix Applied:

✅ Enhanced patched-start.sh to replace API URLs at runtime
✅ Added debug endpoint /api/v1/debug/deployment for verification
✅ Added deployment verification script verify-isolation.sh

Next Debugging Steps:

Deploy the updated configuration
Check container logs for any startup errors

Run debug script inside container:

docker exec -it <container_name> /app/debug-services.sh

Test internal services:

# Test frontend
curl http://localhost:8000
# Test backend  
curl http://localhost:9000
# Test nginx
curl http://localhost:80

Troubleshooting Commands:

# Check PM2 processes
docker exec -it <container> pm2 list

# Check ports in use
docker exec -it <container> netstat -tlnp

# Check nginx config
docker exec -it <container> nginx -t

# View PM2 logs
docker exec -it <container> pm2 logs

# Run full debug
docker exec -it <container> /app/debug-services.sh

Expected Resolution:

The 502 errors should resolve once:

✅ Frontend service starts correctly on port 8000 (WORKING)
❌ Backend service starts correctly on port 9000 (FIXED - needs redeploy)
✅ Nginx properly proxies requests between them (WORKING)

Post-Deploy Verification:

After redeploying, verify isolation works:

# Run the automated verification script
./verify-isolation.sh

# Or manually test the debug endpoints
curl https://adr-lms.whitex.cloud/api/v1/debug/deployment
curl https://edu.adradviser.ro/api/v1/debug/deployment

# Check for cross-deployment API calls in browser Network tab
# Should see only same-domain API calls:
# - DEV: Only calls to adr-lms.whitex.cloud
# - LIVE: Only calls to edu.adradviser.ro

Expected output should show:

✅ Different database hosts for DEV vs LIVE
✅ Different cookie domains: adr-lms.whitex.cloud vs edu.adradviser.ro
✅ No cross-domain API calls in browser Network tab
✅ Separate content/courses on each deployment

The port configuration fix was the critical missing piece for Traefik routing.

4.3 KiB Raw Blame History