- Idempotent (Re-run of Stack creation) : Currently you can only create/delete a stack. If part of the create fails, default is to rollback all. You can adjust that, but I found no way to re-run the create so that it would create the missing pieces. Also if you destroy an instance by hand, this would come in handy to rebuild part of the stack
- Slowness: the creation process on paper look as if the dependencies are handled intelligently and possible speedier. I found creating individual instances in parallel faster than creating them in a stack
- Throttling limit: if you delete, create stack again (f.i. because it failed), I often get errors for throtling limit. I think this limit is not imposed on the regular creation of instances.
- If a creation fails, you have to delete the stack first. Maybe you need another state, so you don't need to delete it first (like state-terminated)
- Name limit: this only accepts alphanum , no spaces or dashes to make it more readable in the console.
- Region Dependency: it seems that you can't create EC::Instances from AMI's that are not in the same region as where your stack is created.
- Better checking upfront during the create: f.i. check if keys exist or ami's exist could save you from spinning up 4 other instances and seeing them being destroyed because of a minor error. This is quite costly for debugging.
- Have tags for a cloudformation? Such as tags["Name"]
- Standard place for user to access machine: when automating logins you have to guess this, is it root, ubuntu, ec2user
- A way to sync keys/AMI's across regions : instead of having to create them/manage them in each center individually