Provisioning stability

[ehazlett]

The creation process varies greatly from provider to provider. Issues like network partitions, provider API hiccups, etc can cause issues. Currently, Machine does not handle those very well. Errors are not always reported well and the user must remove whatever has been created and start over. This affects users using Machine directly as well as indirectly (i.e. Kitematic, Rancher, Machinery). We should improve the provision process to handle various types of failure and better report them when they occur.

Related:

• Output from SSH command that fails should be printed #1266
• create fails due to: No IP address found #721
• docker-machine hangs on create #1028
• Random problem downloading docker #950
• Another reason docker-machine create is hanging #1031
• Re-provision host without needing a re-create #1330
• machine 0.3 does not work if docker daemon needs http_proxy #1319

[dgageot]

Closing. Since only #1330 is still opened