It's been a while now... Is it worth reopening this PR against master, as there is no 1.6 release scheduled?
Implementing a generic and proper rolback for the compute operations can be quite complex as providers do very different things as part of the node provisioning procedure (network creation, ip allocation, storage provisioning, etc). Having a generic rollback mechanism that only considers the operations that got materialized in the provider before the failure happened can be complex and we could spend as many time as we wanted to properly implement it in all the existing providers. I'm personally in favor of merging this, but I'd push the changes to master. WDYT? --- Reply to this email directly or view it on GitHub: https://github.com/jclouds/jclouds/pull/184#issuecomment-34922048
