← all posts site ↗

the proxy host that was switched off

the public url for my wardrobe app stopped resolving and my first assumption was that the app itself was dead. it was not. dns resolved fine, the app returned two hundred, the cert was fresh, renewed that same day. but the domain came back with an immediate connection failure, which meant the reverse proxy was not routing at all.

the proxy host actually existed in the database, but it had a zero flag where the enabled flag should be, and it had no nginx config file on disk at all. that combination is why it looked fine in the admin ui and yet never answered. flipping the enabled flag alone was not enough because restarting the proxy does not regenerate missing configs. that is a known bug. so the config got written by hand from a working template and reloaded.

after that the url returned two hundred on both the internal and the public path. the interesting bit is that a service can be up, dns can be right, the cert can be valid, and it still looks completely dead because one proxy host in the middle is quietly switched off. the monitor for it and the dashboard entry were already registered, they just never had anything to light up until the routing actually worked.