HBase Restserver startup probe failing with external listeners

[xeniape]

#708 revealed in weekly tests that the restserver startup probe is using port 8080 (the default rest port) instead of the one assigned by the listener, same for HBase restserver itself which before #708 used the default port 8080 instead of getting assigned the listener port. Maybe other usages of that as well. Subsequently the Pod never gets ready.

Goal would be to analyze and fix the problems.

Description about the analyzed problem and the solution are found in #716

Acceptance

• Passing integrationtests https://testing.stackable.tech/view/02%20Operator%20Tests%20(custom)/job/hbase-operator-it-custom/72/

[razvan]

One related question is why are REST servers now special cased?

The original issue wanted them advertised. We decided to not do that as suggested (add them to the discovery map). Instead, every REST server Pod now has it's own listener service while the other roles (master and regionserver) do not.

IMO especially for REST servers this is too much. Clients never need to work with all or with a specific instance of them. These are not data bearing servers. There is no value added in having a service per http(s) Pod.

[xeniape]

Instead, every REST server Pod now has it's own listener service while the other roles (master and regionserver) do not.

Can't comment on any of your other points, but at least for this one: master and regionserver Pods also have a listener service per Pod. Not sure if that changes anything.

[razvan]

Pods also have a listener service per Pod.

OMG! I was confused by this. But of course there is also this

[xeniape]

Release Note in #716 (comment)