Improve CassandraHealthIndicator with more robust mechanism

[tomekl007]

…22901

[adutra]

I'm still seeing lots of checkstyle errors, could you please run ./gradlew formatMain formatTest checkstyleMain checkstyleTest and fix the errors?

[adutra]

I actually prefer the non-reactive version that I suggested because it is simpler and easier to override (you only need to extract the Metadata object, without caring about wrapping the result in a Mono).

[tomekl007]

I also prefer it but I think that is the point of this CassandraReactiveHealthIndicator.java.
We should use .getReactiveCqlOperations().execute and it imposes that it needs to be wrapped in the Mono

[adutra]

1. Then you should do the same for buildHealth – why is it not returning Mono<Health>?
2. Make this method protected please.

[tomekl007]

1. buildHealth is invoked in the map() on Flux therefore it does not need to return Mono (and shouldn't)
   ReactiveSessionCallback has Publisher<T> doInSession so it must return Flux or Mono. Why do you want to make the buildHealth return mono?

[adutra]

I don't understand your logic. You told me that "the point of this CassandraReactiveHealthIndicator.java" is that all methods should return reactive types, but now you say that this method should not return a reactive type? Anyways I don't care, it's fine the way it is. Stéphane is going to rewrite everything to his taste anyways :-)

[tomekl007]

the problem and limitation is that we should use reactiveCassandraOperations.getReactiveCqlOperations().execute that returns the Flux so we must return reactive type from extractMetadata(), next we are just chaining operations on the Flux so I see no reason to change the buildHealth to return reactive type. I will open the public PR in that case.

[tomekl007]

I'm still seeing lots of checkstyle errors, could you please run ./gradlew formatMain formatTest checkstyleMain checkstyleTest and fix the errors?

done