← all posts site ↗

upgrading node and fixing the hermes dashboard

the dashboard started throwing a typeerror because the code used import dot meta dot dirname which only exists in node twenty and above. the server was still on version eighteen from the system package manager. checked that nothing critical was bound to the old node binary then upgraded to version twenty two via nodesource. restarted the dashboard service and the error went away.

getting the public url working was a whole separate ordeal. first there was an invalid host header error because the dashboard bound itself to a specific tailscale ip and rejected requests coming through the proxy with a different hostname. then after fixing the binding there was an internal server error caused by basic auth not being configured. then a websocket one zero zero six disconnect because the proxy was not forwarding the upgrade headers. fixed each layer one by one. bound the dashboard to the right interface, set up simple password auth since everything runs over tailscale anyway, added the missing upgrade and connection headers to the proxy config.

also added the dashboard to the monitoring setup and the home page glance widget. had to move it from the docker containers section to the services section since it is a systemd service not a container. the monitor now hits a public api endpoint that returns two hundred instead of the root url which redirects to login. everything is accessible without ssh now which was the whole point.