Roll Out Framework with Cluster-Formation

[freeformflow]

These edits move PandaCluster beyond simply creating a CoreOS cluster. The goal is to setup a deployment environment with a single command. It should be noted that the following abilities apply only to cluster formation. This is in line with instructions to focus on making clusters ephemeral and disposable. If we wish to make changes to a running server, it would take more effort.

Now you can create a cluster with:

1. Arbitrary public SSH keys loaded on every machine.
2. SkyDNS deployed on every machine. (This pulls the image, but needs some tweeks)
3. A hook-server deployed. (This pulls the image, but the service still fails utterly)

   Changes

   • Build-Template command has been removed
   • public_keys and hook_server fields are now located in the PandaCluster dotfile to configure the cluster.
   • The units directory has been moved into src/services. SkyDNS and the hook-server service files have been added to the library
   • Services are launched during cluster-formation by selecting from among those listed in the units.cson manifest, much like how arguments are specified.

     ToDos

   • I need to work the bugs out of SkyDNS and hook-server services in the cluster. SkyDNS appears to basically work, but there is something wrong with the hook-server still.
   • I need to work out a graceful way to edit the *.service files. I want to turn them into templates we can manipulate with CoffeeScript. This should make them much more dynamic that systemd is by intself, haha.
   • There needs to be a function to allow the user to select a URL for their cluster.
   • None of this is documented yet, so I'll need to do that before I get hit by a truck. :truck:

[freeformflow]

This is stable and addresses #3, #4, and #5. @peterlnguyen, we should talk about merging this tomorrow with your approval.

[freeformflow]

This is no longer relevant because of the branch feature/private-dns