Gatus is a project that I started because I wanted to have an automated service health dashboard, and I wanted it to be as low maintenance as possible.
Originally, determining whether a service was healthy or not was going to be very simple: is the status 5xx? If yes, the service is unhealthy, otherwise, it's healthy.
After implementing that, however, I felt that it was bit too lacking. Sure, it could've helped roughly determining the availability of a service, but it wasn't able to make the difference between "everything is ok" and "everything is not ok".
For instance, let's say you want to test an endpoint that should return a list of recipes, and for whatever reason, the endpoint starts returning an empty list of recipes. The endpoint would still return a 200, but there's no content, so is it really accurate to call the service "healthy"?
That's why I implemented a way to supply conditions that Gatus will evaluate in order to determine whether a service is healthy or not. These conditions range from status code to body comparison with the help of a JSONPath-like syntax. I also extended that by adding the ability to resolve the length of the value of said JSON path, which allows even greater flexibility for when you don't know exactly what the data is (or the data is bound to change), but you know that this data should exist, or a certain number of elements should be returned.
For the example I mentioned earlier, the solution would simply be to add the following condition to the service configuration: `len([BODY].recipes) > 0`. This condition would make sure that the response body has the field `recipes` and it has a length of at least 1.
Originally, determining whether a service was healthy or not was going to be very simple: is the status 5xx? If yes, the service is unhealthy, otherwise, it's healthy.
After implementing that, however, I felt that it was bit too lacking. Sure, it could've helped roughly determining the availability of a service, but it wasn't able to make the difference between "everything is ok" and "everything is not ok".
For instance, let's say you want to test an endpoint that should return a list of recipes, and for whatever reason, the endpoint starts returning an empty list of recipes. The endpoint would still return a 200, but there's no content, so is it really accurate to call the service "healthy"?
That's why I implemented a way to supply conditions that Gatus will evaluate in order to determine whether a service is healthy or not. These conditions range from status code to body comparison with the help of a JSONPath-like syntax. I also extended that by adding the ability to resolve the length of the value of said JSON path, which allows even greater flexibility for when you don't know exactly what the data is (or the data is bound to change), but you know that this data should exist, or a certain number of elements should be returned.
For the example I mentioned earlier, the solution would simply be to add the following condition to the service configuration: `len([BODY].recipes) > 0`. This condition would make sure that the response body has the field `recipes` and it has a length of at least 1.
I wrote a short article on my website, if you want a bit more context: https://twinnation.org/articles/46/gatus-automated-health-da...
Any feedback is greatly appreciated :)