Suggested Config Fix: 502 Bad Gateway on /dashboard when using Nginx Proxy Manager on a separate server

[UncleDoomVSSP]

Fix: 502 Bad Gateway on /dashboard when using Nginx Proxy Manager on a separate server

Symptom: The login page loads fine, but after logging in you get a 502 Bad Gateway (openresty) on /dashboard and other internal routes. Everything looks correctly configured and direct access to the container port works fine.

Cause: SvelteKit sends a large number of link preload headers during server-side rendering (SSR). Nginx's default proxy buffer size is too small to handle them, so NPM returns a 502 before the response reaches the browser.

You can confirm this by checking your NPM proxy host error log:

docker exec -it <npm-container-name> cat /data/logs/proxy-host-XX_error.log

Look for this line:

upstream sent too big header while reading response header from upstream

Fix: In NPM, edit the proxy host for your frontend domain, go to the Advanced tab, and add:

proxy_busy_buffers_size 512k;
proxy_buffers 4 512k;
proxy_buffer_size 256k;

Save and the 502 should resolve immediately.

Also worth noting for NPM users:

• PUBLIC_SERVER_URL must stay as http://server:8000 (the internal Docker hostname). Do not change it to your public backend URL, even though the error page hints to check it.
• Make sure Websockets Support is enabled on the frontend proxy host.
• The NPM "Online" status indicator only checks TCP connectivity, not whether requests are actually being forwarded successfully, so it can show green even when 502s are occurring.

Hope this saves someone a few hours!

[lovelacecreativeco]

Thank you for this! Immediately solved my problem.