Is there a way to fail fast on deploy if first instance deployment fails

[pbrumm]

i have a db box setup that runs migrations. if migrations fail, I don't want app boxes to get deployed to.

it is not obvious how I could stop deployment on error.

[rick]

Unfortunately there's probably not an easy way to make this happen. Each deployment is linear - i.e., every step happens in order for each individual host. As opposed to being jointly sequential (e.g., as capistrano or vlad would do) -- where each phase of the deployment runs on all hosts, and if it fails, then the next phase is stopped for all hosts.

I had looked at doing some work on synchronization to overcome this limitation, but had never gotten back to it. The environments where I do all my deployment these days end up doing migrations out-of-band anyway (we branch separate commits for migrations and deploy them separately in advance of any usage), so I'm not even dogfooding this. :-/