Appearance
TLS and reverse proxy
A Substrate node does not speak TLS, does not rate-limit by origin, and should not expose 9944 to the internet. HTTP and WebSocket share one port (9944). Port 9933 does not exist in modern Substrate. A location = /rpc in nginx that proxy_passes 9933 fails in the worst way: exact-match wins over prefix, so the URL you put in NEXT_PUBLIC_RPC_URL is the dead one, while WS still works.
What is actually reachable
| Host | TLS | Serves the chain |
|---|---|---|
https://rpc2.testnet.empooriochain.com | Let’s Encrypt | yes — HTTP JSON-RPC and wss:// |
rpc.testnet.empooriochain.com | nginx CN=invalid, return 444 | no |
| Cleartext 9944 / 9946 | closed 2026-10-27 | no |
Validator 1 still authors. It is reachable only under a .org name whose zone died 2026-10-11 (curl --resolve diagnostics). The fix is add rpc.testnet.empooriochain.com to the working nginx server_name and certbot --nginx -d rpc.testnet.empooriochain.com. Until then the testnet has one public door.
P2P is TCP 8443 / 8447, not through the proxy. Prometheus stays internal.
Nginx sketch
Rate-limit at the proxy. The node has none of its own. Two validators and zero fault tolerance make a single noisy IP a liveness problem.
nginx
limit_req_zone $binary_remote_addr zone=rpc_req:10m rate=20r/s;
server {
listen 443 ssl http2;
server_name rpc2.testnet.empooriochain.com;
location / {
limit_req zone=rpc_req burst=40 nodelay;
proxy_pass http://127.0.0.1:9944;
proxy_http_version 1.1;
proxy_set_header Upgrade $http_upgrade;
proxy_set_header Connection "upgrade";
proxy_set_header Host $host;
proxy_read_timeout 3600s;
}
}Do not use location = /rpc unless that path also hits 9944.
Marketing and docs hosting are Firebase. This page is for the nodes, not for empooriochain.com.
CORS
The public RPC sends access-control-allow-origin: *. Browser dApps and the marketing hero depend on it.
Safe vs unsafe
Public RPC: Safe. author_rotateKeys rejected. Session-key generation: localhost + --rpc-methods unsafe only.
