5.1 KiB
KoalaSync Relay Server
A Node.js relay server for synchronized video playback.
Key Features
- Zero-Persistence: No database. All state is held in RAM.
- Privacy First: No tracking, no persistent logging of user data.
- WebSocket Only: Minimal overhead with efficient transport.
Setup
Environment
Copy .env.example to .env and configure your settings.
PORT="3000"
MAX_ROOMS="1000"
MAX_PEERS_PER_ROOM="25"
MIN_VERSION="1.0.0"
# Required for production/self-hosting: unique random salt used for room-password hashes.
# Generate one with: openssl rand -base64 32
SERVER_SALT=""
# Optional: enables aggregate-only admin metrics on /health with Authorization: Bearer <token>
# Use a long random token, 32+ characters recommended.
ADMIN_METRICS_TOKEN=""
# Optional: set to "1" to enable verbose connection, room-join/leave, and CORS logs in the console. Default is "0" (disabled).
DEBUG_LOGGING="0"
Room Password Salt
SERVER_SALT is used to HMAC room passwords before they are stored in RAM. The relay has a built-in fallback so development keeps working, but that fallback is public in the repository and the server logs a startup warning when it is used.
For every real self-hosted deployment, set a unique random value:
openssl rand -base64 32
Then add it to server/.env or your Compose environment:
SERVER_SALT=replace-with-a-long-random-salt
Health & Metrics
GET / and GET /health are IP-rate-limited to 10 requests per minute per client IP. These health-style responses are cached server-side for 60 seconds and refreshed lazily on request. By default /health returns only basic service status, uptime, room count, connection count, and a timestamp.
If ADMIN_METRICS_TOKEN is set, requests with Authorization: Bearer <token> receive additional aggregate metrics such as total peers, average peers per room, max room size, active lobby count, rate-limit map sizes, and process memory usage. Wrong admin bearer attempts are separately limited to 5 requests per minute per client IP. The metrics response does not include room IDs, peer IDs, usernames, IP addresses, media titles, or other user-level data.
Use a long random ADMIN_METRICS_TOKEN of at least 32 characters. Shorter configured tokens still work, but the server logs a startup warning.
Generate a token with one of these commands:
openssl rand -base64 32
# or
node -e "console.log(require('crypto').randomBytes(32).toString('base64'))"
For Docker Compose deployments, set the generated value in server/.env:
ADMIN_METRICS_TOKEN=replace-with-a-long-random-token
When polling /health from Prometheus, Uptime Kuma, cron, or a load balancer, keep the interval comfortably below the public limit of 10 requests per minute per client IP. A 30-60 second interval is recommended for routine monitoring. Use the admin bearer token only from trusted monitoring hosts, and keep the Node server private behind Caddy or another trusted reverse proxy because IP-based limits depend on the configured proxy boundary.
Docker (Recommended)
The server is available as a pre-built image on GHCR.
# Pull from GHCR
docker pull ghcr.io/shik3i/koalasync:latest
# Or build from the repository root
docker build -t koala-sync-server -f server/Dockerfile .
See Docker network compose or Static IP compose in the root directory for ready-to-use Docker Compose files.
Manual Setup
cd server
npm install
npm start
Verification
The server is covered by the root verification suite. From the repository root, run:
npm run verify
For focused server checks, see scripts/test-server-ops.mjs, scripts/test-server-routes.mjs, scripts/test-server-ws.mjs, and scripts/test-rate-limiter.mjs.
Security
- Rate Limiting: IP-based connection limits and socket-based event limits.
- Health Endpoint Throttle:
GET /andGET /healthare limited to 10 requests per minute per IP, with 60-second lazy server-side response caching and stricter throttling for wrong admin bearer attempts. - Room Discovery Throttle: Room-list refreshes are rate-limited server-side to one request every 10 seconds per socket.
- Token Handshake: Requires a valid token defined in the root
shared/constants.js. - Password Hash Salt: Set
SERVER_SALTin every deployment so room-password hashes are not derived with the public fallback salt. - Single Source of Truth: The server imports constants directly from the root
shared/directory. - In-Memory: Rooms are automatically pruned after 2 hours of inactivity.
- Reverse Proxy Boundary: The server trusts one reverse proxy hop for client IP detection. Keep the Node port private/firewalled so clients can only reach it through Caddy or another trusted proxy.
Module Structure
| File | Purpose |
|---|---|
index.js |
Express + Socket.IO server: room management, relay loop, graceful shutdown |
rate-limiter.js |
Connection, event, health, and auth rate limiting with 6 functions + cleanup intervals |
ops.js |
Health endpoint helpers, metrics payload builder, auth validation |