Health check API

Merkmal
1

It would be nice to have some sort of health check API. We recently faced an issue with an update and Discourse was producing 500 errors.

However, curl returned 200:

curl -I https://forum.gitlab.com
HTTP/2 200
server: nginx
2 „Gefällt mir“
2

This exists, see

/srv/status

14 „Gefällt mir“
3

Ah, thank you! I was searching “health check” and that didn’t yield any results.

2 „Gefällt mir“
4

Post is 404. What are the endpoints?

5

I’ve updated my post.

2 „Gefällt mir“
6

Actually I don’t think /srv/status would catch migration issues like the one mentioned above…

(and it would be pretty hard to build a check that does catch issues like that one)

2 „Gefällt mir“
7

Yes… /srv/status is there as a very cheap test, all it does is ensures the apps middleware stack is working.

To catch issues where you auto deploy I would recommend monitoring 200s, if there is a large increase in non 200s alert.

5 „Gefällt mir“
8

Would https://discourse.example.org/srv/status be a good place to point an uptime monitor? I am thinking it may not be enough to have a reliable measure of “is the site up”, but it would be nice to have something that produces less load on the system for monitoring purposes.

(Alternatively, might there be any plans to expand the components listed on this endpoint?)

3 „Gefällt mir“
9

Yeah that is a reasonable spot, you could also point at a specific topic and search for text if you want something more fancy

7 „Gefällt mir“
10

Yeah, we had been using /about but inclined to use this instead.

My old ops/on-call mind woke up to make me think it could still be interesting (and helpful occasionally for troubleshooting) if it were something like:

db ok
middleware ok
whatever-else ok
...
all systems ok
4 „Gefällt mir“