To avoid any kind of downtime of our environment, we can perform a rolling update. A rolling update replaces instances in a particular order. By default, we can choose to either replace one instance at a time or replace half at a time, but we can also customize the replacement order if required. The rolling update will replace the instances with newer instances in the same autoscaling group so that, at a certain point, there will be different versions (the old and the new) running and accepting calls from clients, which might cause issues in certain applications.