Currently log retrieval is done in the same method that runs the tests. If a test fails due to, for example, failure to provision or update the cluster, the resulting exception short-circuits the rest of the method and log retrieval isn't done.

In order to always retrieve server logs on failures, the log retrieval needs to be separate from the test execution method.

This separation will also make failures in log retrieval not affect overall test outcome, which will improve the success rate of the tests overall.