AWS Amazon auto scaling with mixed types of instance

Question

If I have 2 different types of instances say micro and large in my scaling group. Then how does auto scaling work if I have configured min-machines as 2 and max as 10 ? Does it create 5 instance each of micro and large when fully scaled out ?

Answer 1

The Auto Scaling Group will always create new instances according to the Launch Configuration associated with the group. I'm pretty sure that it will scale down by first terminating instances that don't match the current Launch Configuration of the group.

If you want to run an application on a cluster of servers that is made up of two different instance types you can do it by creating two Launch Configurations with two Auto Scaling Groups. If your application requires instances to be registered with an Elastic Load Balancer, you can configure both Auto Scaling Groups to register their instances with the same ELB. This way you can scale the different instance types independently of each other.

Answer 2

Only one type of instance can be on a Launch Configuration. And only one Launch Configuration can be in an Auto Scaling Group. In that way, if you have selected micro instance in the Launch Config, you can see only that type in your Auto Scaling Group. And in that auto scaling group, you must have defined a AutoScalingPolicy which can be 'add 1 instance' or 'add n instance' or 'add n% instance'. Based on that value it will create instance.

For example, if you selected min as 2 and max as 10, and scaling policy as add 1 instance. Initially, there would be 2 instance created and running as part of Auto policy, and every time the threshold exceeds the policy, 1 instance will be created up to 10. No more instances will be created if the threshold goes up after having 10 instances.