Replace Auto Scaling instances based on an instance refresh

As part of starting an instance refresh, you specify the minimum healthy percentage to maintain at all times. This is the amount of capacity in an Auto Scaling group that must pass your health checks during an instance refresh so that the refresh can continue. For example, if the minimum healthy percentage is 90 percent, then 10 percent is the percentage of capacity that will be terminated and replaced. If the new instances don't pass their health checks, Amazon EC2 Auto Scaling terminates and replaces them. If it cannot launch any healthy instances, the instance refresh will eventually fail, leaving the other 90 percent of the group untouched. If the new instances stay healthy and have finished their warm-up period, Amazon EC2 Auto Scaling can continue to replace other instances.

An instance refresh can replace instances one at a time, several at a time, or all at once. To replace one instance at a time, set a minimum healthy percentage of 100 percent. To replace all at once, set a minimum healthy percentage of 0 percent.

Amazon EC2 Auto Scaling determines whether an instance is healthy based on the status of the health checks that your Auto Scaling group uses. For more information, see Health checks for Auto Scaling instances. To make sure that Amazon EC2 Auto Scaling can detect instance health issues as soon as possible, don't set the group's health check grace period too high. For more information, see Set the health check grace period for an Auto Scaling group.

The instance warmup is the time period from when a new instance's state changes to InService to when it can receive traffic. During an instance refresh, if the instances pass their health checks, Amazon EC2 Auto Scaling does not immediately move on to replacing the next instance after determining that a newly launched instance is healthy. It waits for the warm-up period before it moves on to replacing the next instance. This can be helpful when your application takes time to initialize itself before it starts to serve traffic.

You can reduce the value of the warm-up period if you use a lifecycle hook to prepare new instances for use. For more information, see Amazon EC2 Auto Scaling lifecycle hooks.

The desired configuration is the new configuration that you want Amazon EC2 Auto Scaling to deploy across your Auto Scaling group. For example, you can specify a new launch template and new instance types for your instances. During an instance refresh, Amazon EC2 Auto Scaling updates the Auto Scaling group to the desired configuration. If a scale-out event occurs during an instance refresh, Amazon EC2 Auto Scaling launches new instances with the desired configuration instead of the group's current settings. After the instance refresh succeeds, Amazon EC2 Auto Scaling updates the Auto Scaling group settings to reflect the new desired configuration that you specified as part of the instance refresh.

Skip matching tells Amazon EC2 Auto Scaling to ignore instances that already have your latest updates. This way, you don't replace more instances than you need to. This is helpful when you want to make sure your Auto Scaling group uses a particular version of your launch template and only replaces those instances that use a different version.

A checkpoint is a point in time where the instance refresh pauses for a specified amount of time. An instance refresh can contain multiple checkpoints. Amazon EC2 Auto Scaling emits events for each checkpoint. Therefore, you can add an EventBridge rule to send the events to a target, such as Amazon SNS, to be notified when a checkpoint is reached. After a checkpoint is reached, you have the opportunity to verify your deployment. If any problems are identified, you can cancel the instance refresh or roll it back. The ability to deploy updates in phases is a key benefit of checkpoints. If you don't use checkpoints, rolling replacements are performed continuously.