Comment 1 for bug 1313785

I am not able to reproduce this.

juju ssh will now ssh to machine 0's public address directly, and then proxy through that to machine 0's internal address. This is to support environments where machines do not have unique public addresses (i.e. with load balancing).

Can you please ssh to the machine directly and then check whether you can resolve ip-172-31-21-242.us-west-2.compute.internal. If that does not resolve, that would indicate a problem with EC2. If it *does* resolve, then please try "juju ssh" again with --debug and attach the output.