Troubleshooting guide

Nginx Proxy Manager Upstream Timeout: Diagnose a Slow but Alive Backend

Category: Reverse Proxy & SSL

You are running Nginx Proxy Manager in front of a self-hosted service, and requests hang until the proxy returns a gateway timeout even though the backend is still running. This is a classic nginx proxy manager upstream timeout, and it usually means the proxy gave up before your slow service could answer.

Confirm the Upstream Is Slow, Not Dead

First, test the backend directly from the NPM host, bypassing the proxy. Run a curl with a generous timeout:

curl -v --max-time 60 http://<upstream-ip>:<port>/

If curl eventually returns a response, the upstream is alive but slow. If it times out too, the problem is inside the service, not NPM. Also check the NPM container logs for the exact error: upstream timed out (110: Connection timed out) while reading response header from upstream confirms the proxy deadline was hit.

Inspect the Proxy Timeout Settings

In NPM, open the proxy host and look at the advanced configuration. The default proxy_read_timeout is 60 seconds. If your service needs longer, add:

proxy_read_timeout 300s;
proxy_connect_timeout 60s;
proxy_send_timeout 300s;

Save and let NPM reload. Interpretation: if the timeout error disappears but responses are still slow, you have fixed the symptom and can now profile the backend.

Verify the Fix

Repeat the same request through the proxy hostname with curl -v --max-time 300 https://your.domain/. A successful response with a 200 status and no timeout line in the NPM logs means the proxy now waits long enough. Test twice to rule out caching.

Roll Back if Needed

If the longer timeout causes hung connections or memory pressure, remove the custom lines and save. NPM reverts to defaults immediately. For persistent tuning,

A laptop displaying proxy host settings and a browser HTTPS test

Ask for help before changing more directives.


Back to the service page · Ask about this exact symptom