cluster ssh not working for instance in VPC

[gwilton]

I am now able to launch instances inside a VPC and automatically associate an EIP to the instance with version 4.7.2. However, knife "cluster launch" gets stuck in the final phase of the launch while trying to ssh...

sandbox-raid-0: setting termination flag false
sandbox-raid-0: associating Elastic IP 107.23.163.84
sandbox-raid-0: syncing EBS volumes
sandbox-raid-0: launching
sandbox-raid-0: waiting for ready
sandbox-raid-0: trying ssh

After waiting a few minutes and noticing the instance is online with the proper EIP I CTRL+C from the "cluster launch".

I then try to "cluster ssh" to the available instance and get the following...

[homebase (master)]$ knife cluster ssh sandbox uptime
Inventorying servers in sandbox cluster, all facets, all servers
sandbox: Loading chef
sandbox: Loading ec2
sandbox: Reconciling DSL and provider information
FATAL: No nodes returned from search!

It seems the issue lies in lib/chef/knife/cluster_ssh.rb:54. The c.machine.public_hostname is being used for cluster ssh. AWS forum explain that EIPs in VPC don't have a public_hostname set (https://forums.aws.amazon.com/message.jspa?messageID=406306).

        target = target.select {|t| not t.bogus? }
        addresses = target.map {|c| c.machine.public_hostname }.compact

        (ui.fatal("No nodes returned from search!"); exit 10) if addresses.nil? || addresses.length == 0