Allow to not log ping requests

[johanfleury]

Describe the problem to be solved

It can happen that a process crashes in a way where it gets stuck but doesn’t exits. For this reason Docker (and Kubernetes) has a feature called health check that allows to check that the process is still working as expected (and restart the container if not).

This can be done by a script doing some hard work, but most of the time it’s just about curl’ing a simple API endpoint that replies with HTTP status code 200.

PeerTube doesn’t have such an health check endpoint.

Describe the solution you would like:

This endpoint should do the following:

• Reply with status code 200
• (Optional) Reply with a JSON body describing the current status (it’s often just {"status": "ok"})
• Avoid logging request either at the discretion of the administrator (with a specific setting in the configuration files) or at the discretion of the client (with a query parameter e.g. <endpoint>?silent=true)

The path of the endpoint could be something like /health/live (in line with how Kubernetes names this feature, and would allow for an eventual future implementation of a “readiness” use this feature and allow a /health/ready

Describe alternatives you have considered

I am currently using / to check if PeerTube is still alive, but the main issue is that it logs every single request:

$ k logs -n peertube peertube-0 peertube | grep -m 1 kube-probe
[***:443] 2021-01-07 14:38:57.755 info: 10.2.0.1 - - [07/Jan/2021:14:38:57 +0000] "GET / HTTP/1.1" 200 3997 "-" "kube-probe/1.20"
$ k logs -n peertube peertube-0 peertube | grep -c kube-probe
1048

[kontrollanten]

Hi! Have you seen that /api/v1/ping exists? But I think it still logs the requests.

[johanfleury]

Hey, no I didn’t saw it.

Regarding logging, I’ve looked at morgan’s documentation and I think I can come up with a PR.