Potential misuse of gRPC UNAVAILABLE in kuksa.val.v2

[rafaeling]

According to the gRPC status code documentation the UNAVAILABLE status should be used when the gRPC service itself is not available (e.g., server down or unreachable).

In Databroker kuksa.val.v2, however, UNAVAILABLE is currently also used in a different context:
// UNAVAILABLE if there is no provider currently providing the actuator

This might cause confusion, since UNAVAILABLE is intended to describe service-level availability rather than application-level conditions.

Suggestion: We should review this usage and consider whether a different status code (e.g., NOT_FOUND, FAILED_PRECONDITION, or another more appropriate one) would better capture the “no provider for actuator” case

[BjoernAtBosch]

Reading the documentation of FAILED_PRECONDITION, UNAVAILABLE seems to fit best. But I see the potential issue for clients to distinguish between the situations "databroker not avail" and "provider not avail".

Basically, my favorite solution would have been (and still is) not to signal issues on the application level via gRPC status codes but instead define own error signalling in the method responses. This would also help to make the databroker interface independent of the used service technology (i.e. gRPC currently). But of course this would most probably mean another incompatible API change.

Maybe we could stay with UNAVAILABLE but define a definite text message in the status ...