When checking for deployed nodes, the kubernetes join check was
only performing a negative check - and would wait for up to the timeout
even in the event that nodes that were not part of the current
processing before proceeding. This had the drawback of being overall
likely to add wait time in any complex deployment scenario, as well as
(and more importantly) miss the case where a node never started to try
to join, and assume that was a success.
This patchset flips the logic to positively look for an expected set of
nodes instead, and will not wait upon nodes that are not currently being
checked. The end result should remedy both of the drawbacks listed
above.
Change-Id: Ib07e4e2677ec4f773d695d57893fdfa5e4b7ff76