How do I get AWS CloudFormation to recreate a resource I manually deleted?

I'm experimenting with CloudFormation right now. I deployed a stack that included an AutoScalingGroup (and LaunchConfig and CloudFormation::Init), LoadBalancer, a single EC2 instance (and CloudFormation::Init), and a couple security groups.

If I terminate EC2 instances created by the AutoScalingGroup, they get recreated as expected. This is actually quite useful for testing changes to my LaunchConfiguration and CloudFormation::Init settings (basically making sure it works from scratch).

I also have an EC2 instance (named DnsServer1) that has a public IP. While working on its CloudFormation::Init data, I also manually terminated its instance. I know CloudFormation doesn't restore it like an AutoScalingGroup does, but I was expecting it would be recreated when I ran aws cloudformation update-stack.

Instead, I get the error: AWS::EC2::Instance DnsServer1 i-014eee8720c4fb542 does not exist.

Full update log:

11:03:40 UTC-0500   UPDATE_ROLLBACK_COMPLETE    AWS::CloudFormation::Stack  GregCFTest1 
11:03:38 UTC-0500   UPDATE_ROLLBACK_COMPLETE_CLEANUP_IN_PROGRESS    AWS::CloudFormation::Stack  GregCFTest1 
11:03:36 UTC-0500   UPDATE_COMPLETE AWS::AutoScaling::LaunchConfiguration   WebServerLaunchConfig   
11:03:36 UTC-0500   UPDATE_COMPLETE AWS::EC2::Instance  DnsServer1  
11:03:13 UTC-0500   UPDATE_ROLLBACK_IN_PROGRESS AWS::CloudFormation::Stack  GregCFTest1 The following resource(s) failed to update: [DnsServer1].
11:03:11 UTC-0500   UPDATE_COMPLETE AWS::AutoScaling::LaunchConfiguration   WebServerLaunchConfig   
11:03:10 UTC-0500   UPDATE_FAILED   AWS::EC2::Instance  DnsServer1  i-014eee8720c4fb542 does not exist
11:03:00 UTC-0500   UPDATE_IN_PROGRESS  AWS::CloudFormation::Stack  GregCFTest1 User Initiated

What can I do to get CloudFormation to recreate this resource, short of deleting the stack and recreating it?

The easiest way is to give the EC2 instance a different resource name and update the stack. Cloudformation will interpret that as you creating an entirely new instance and create it.

Another alternative is to change a configuration value that forces the creation of a new instance, such as changing the subnet of the instance.