To have systems up & running is something, but to have them healthy it’s another story.
If you are looking to implement healthchecks in your architecture beware of being a bit too simplistic: you might configure, for example, on of your frontend machines to check the status of the backend ones every few seconds, so that nginx or haproxy can remove the backend if they find it unreliable / unhealthy.
Problem is, there might be tricky situations in which the
backend responds with a
200 Ok even though it’s not
There are a lot of ways to avoid this, but the simplest one,
that takes you 2 minutes and works quite well, is to use a
custom HTTP status code for your healthcheck page – we use
For example, in node, we would do:
1 2 3
and then we would need to tell our backend that the only status
code that has to be considered healthy is
No more, no less.
- For example, the backend’s nginx can just respond with nginx’s default welcome page, if your host is misconfigured ↩