Nginx "Bad gateway" after restarting TinyPilot

There seems to be an issue where nginx will report "

Repro steps

1. Start a TinyPilot device normally
2. sudo service tinypilot restart
3. Visit http://tinypilot/

Expected

TinyPilot web interface loads normally.

Actual

Nginx reports “Bad Gateway”

Workaround

Restart nginx:

sudo service nginx restart

[Bramzor]

“Bad gateway” is actually a valid answer during the reboot and shouldn’t be confused with “Gateway Timeout”. A gateway timeout is that the service is running, but not providing an answer within the specified timeframe. Can be increased as mentioned by somik but this will not solve a “Bad gateway” issue. A Bad gateway happens when Nginx tries to connect to the service but for some reason, that port is not up, blocked or tcp session could not be set up or any other reason.

So I assume Nginx is running before the service is ready? For this, there are a few possible ways to avoid this:

• (On service level) Start the nginx service after starting the end service by specifying an “After” in the .service file.
• (On nginx level) Using something like “health_check interval=10 fails=3 passes=1” in nginx which might help, haven’t tried it myself yet. But this defines how long Nginx will see an origin as bad.
• (On application level) When clicking on reboot, use Ajax requests to check if the backend is available again and than refresh the browser.

[mtlynch]

I’d see lines like this:

2020/08/08 20:57:14 [error] 12922#12922: *95 connect() failed (111: Connection refused) while connecting to upstream, client: 10.0.0.100, server: tinypilot, request: "GET / HTTP/1.1", upstream: "http://[::1]:8000/", host: "tinypilot"

But I confirmed a fix in https://github.com/mtlynch/ansible-role-tinypilot/pull/27