While trying to work out the issues with the reloading classloader, I
noticed that the 2nd tier classloader has a hard-coded path which utilizes
a series of hadoop directories, as well as zookeeper directories. This
causes my concern because, while I'm not sure what exactly this 2nd tier
NEEDS to load to get the VFS stuff going, there is always a possibility for
these jars to migrate between versions/packages and then there would be no
way to support without a code change.
Does anyone know what exactly this tier needs to load, and then can we see
about paring down this list and exposing it as something configurable?