Fails if TELEGRAM_STAT is not set

[noplanman]

For me, using docker, the container fails if the TELEGRAM_STAT variable isn't set.

Maybe because the healthcheck uses it but it's not available?

[MarcoBuster]

Can you send us the logs please?

[noplanman]

Hi @MarcoBuster, which logs?

When the container starts, the status is starting for a long time, then it says unhealthy. I can't connect to the API and simply get a "404 page not found".

The container logs only have a single line with the command for starting the API server, nothing more.

[MarcoBuster]

@noplanman set TELEGRAM_VERBOSITY=4 then the logs will appear. Without them we can't understand the problem.

[luckydonald]

@MarcoBuster Isn't the health check like a curl request to the stats port?

It simply says "curl -f http://localhost:8082/ || exit 1"
In the Dockerfile

Of cause that'll fail if there's no :8082 port serving the stats.
The locks would be the wrong place to look here anyway, and I think it's not needed; that's some simple logic error.
As we now have the ping command, we should use that one instead of the stats endpoint.

[luckydonald]

@luckydonald *logs
(Sorry am on mobile)

[giuseppeM99]

@luckydonald the ping command requires a valid token

[luckydonald]

@luckydonald Wait, we can't use a ping command as that is bound to a bot so we would need to add a bot token?

[luckydonald]

@luckydonald So to fix it properly, we should either so if the ping port without a token (unauthorized) as well.
Maybe for now, if the stats are not explicitly enabled they could be instead be enabled but bound to just the local network interface needed to get the curl request done?

[spontanurlaub]

@luckydonald Theoretically ping doesn't require a token, we could move the ping command out of there, but it is more confusing with the URL layout then.

[MarcoBuster]

...or we could just disable the health check if you don't have TELEGRAM_STAT set

[giuseppeM99]

@MarcoBuster is that even possible?

[MarcoBuster]

@giuseppeM99 I don't know, we should investigate

[giuseppeM99]

@MarcoBuster i'll fix it asap

[luckydonald]

Okey, just to give my two cents, maybe we should actually look into creating a real /healtcheck endpoint.
So if it does a ping or not, would be a implementation detail.
Maybe that could even be served at the root directly, doing a few non-critical stats there (See #37?)