Instance Maintenance Policy & Refresh Behavior

When an Auto Scaling Group's Launch Template is updated (a new AMI, a changed configuration) or an instance refresh is explicitly triggered, the group needs a policy for how to replace existing instances with new ones without creating a capacity gap. The instance maintenance policy controls this by defining minimum and maximum healthy percentage bounds during the replacement process — for example, a minimum healthy percentage of 90% and maximum of 110% means the group briefly launches a small number of extra instances on the new configuration before terminating old ones, ensuring capacity never drops below 90% of desired count at any point during the rollout, at the cost of briefly running slightly more instances (and paying for them) than the steady-state desired count. This is conceptually similar to the deployment strategies used elsewhere on AWS (like ECS rolling updates), applied specifically to EC2 Auto Scaling Groups replacing raw instances rather than containerized tasks.