Fix "no open ports detected" by making the HTTP server listen on host 0.0.0.0 and the port supplied in the PORT environment variable. Localhost and 127.0.0.1 accept traffic only from inside the same container. A hosting proxy cannot reach that loopback address.

The process can be healthy while the site is unreachable. This is a networking contract, not necessarily an application crash.

Two values must be right at the same time

SETTING	LOCAL VALUE THAT OFTEN FAILS	DEPLOYMENT VALUE
Host interface	localhost or 127.0.0.1	0.0.0.0
Port	A hardcoded 3000, 5000, or 8000	The value in PORT

Binding to 0.0.0.0 does not publish the container directly to the internet. It lets the platform proxy reach the process on the container network. The proxy normally accepts public HTTPS, terminates TLS, and forwards the request to your HTTP listener.

Working Node and Python examples

const port = Number(process.env.PORT) || 3000;

app.listen(port, '0.0.0.0', () => {
  console.log(`listening on 0.0.0.0:${port}`);
});

Express reads the assigned port and listens on every container interface.

PYTHON

import os
import uvicorn

port = int(os.environ.get("PORT", "8000"))
uvicorn.run(app, host="0.0.0.0", port=port)

The equivalent Uvicorn start inside a Python entry point.

The fallback keeps local development convenient. In production, the host-provided PORT wins. Do not copy one provider's default port into the code and assume every environment uses it.

Read the first symptom, not the final error page

SYMPTOM	LIKELY CAUSE	NEXT CHECK
No open ports detected	Nothing is listening where the scanner expects	Print the host and port at startup
502 Bad Gateway with a running process	Proxy cannot reach the listener	Check 0.0.0.0, PORT, and startup completion
Connection refused	Process exited or never opened the socket	Read the first runtime error
EADDRINUSE or address already in use	Two servers chose the same port	Start one HTTP listener, not one per module
Health check times out	Route is slow, blocked, or never returns success	Create a small endpoint with no database dependency

Reproduce the hosting contract locally

01
Use the real start commandRun the same npm script, Uvicorn command, or entry file configured for the deployment.
02
Set a temporary PORTStart with PORT=10000 or another unused local port so hardcoded values become visible.
03
Print the listener addressThe startup line should name 0.0.0.0 and the value you supplied.
04
Call a small routeUse curl against localhost on that port. A plain health response separates networking from database or frontend errors.
05
Redeploy and read the earliest linesLater proxy errors are consequences. The startup log usually contains the cause.

Common fixes that create a second problem

Do not bind directly to port 443. The platform proxy owns public HTTPS.
Do not set PORT to a domain name or full URL. It is an integer.
Do not open a second server only for health checks. Put the health route on the existing app.
Do not use a development server in production when the framework provides a production start command.
Do not treat 0.0.0.0 as a browser address. Visit the public host URL or localhost during local testing.

Spocket assigns PORT to HTTP apps and routes an HTTPS spocket.dev address to that listener. Use /documentation/domains-and-webhooks for the exact contract and /documentation/logs-and-status for startup output. Static exports do not need a listener, so /run/sites also helps confirm whether the project is a static site or a server app.

No open ports detected: bind to 0.0.0.0 and PORT

Two values must be right at the same time

Working Node and Python examples

Read the first symptom, not the final error page

Reproduce the hosting contract locally

Common fixes that create a second problem

Need somewhere to put it?

Playwright out of memory: stop Chromium from crashing

Cron job running at the wrong time: fix UTC and DST errors

Telegram bot 409 Conflict: stop duplicate getUpdates requests