We have just been building some EC2 server images for Windows/VS2012 CI servers triggered through bamboo (see http://davidstringeratwork.wordpress.com/2012/10/04/using-atlassian-bamboo-ondemand-for-visual-studio-2012-projects-part-1/). We built an AMI image of the CI server and configured bamboo to launch that when we committed to GITHub.
We encountered a problem when we added some other, non-transient instances to help with specialist testing. At that point, our automated CI builds would no longer trigger. We were helped out by the support folks to identify that the problem related to the "number of elastic instances on your AWS account that are not controlled by Bamboo does not exceed" setting in the bamboo configuration. Once we had some more running instances we had exceeded this value so bamboo would not launch any more instances, even though the instances were for different purposes.
The question is this: should this bamboo configuration setting not just relate to instances started from AMI images are launched by bamboo rather than all instances regardles of the AMI from which they originate? Is there something we are missing here in terms of a problem the current behaviour is solving that we haven't encountered yet?
Thanks
Andy