Launch instances using Capacity Blocks
To use your Capacity Block, you must specify the Capacity Block reservation ID when launching instances. Launching an instance into a Capacity Block reduces the available capacity by the number of instances launched. For example, if your purchased instance capacity is eight instances and you launch four instances, the available capacity is reduced by four.
If you terminate an instance running in the Capacity Block before the reservation ends,
you can launch a new instance in its place. When you stop or terminate an instance
in a Capacity Block, it takes several minutes to clean up your instance before you can
launch another instance to replace it. During this time, your instance will be in a
stopping or shutting-down state. After this process is complete, your
instance state will change to stopped or terminated. Then,
the available capacity in your Capacity Block will update to show another instance
available to use.
Requirements
-
Your instance can't launch in a subnet in a different Availability Zone from the Availability Zone where your Capacity Block is located.
-
Your instance can't launch using an AMI with a different platform than the platform for your Capacity Block.
-
To use
P6e-GB200UltraServer Capacity Blocks, you must be opted in to the Dallas Local Zone (N. Virginia) Local Zone.
Related resources
-
To create a launch template targeting a Capacity Block, see Store instance launch parameters in Amazon EC2 launch templates.
-
To launch instances into a Capacity Block using EC2 Fleet, see Tutorial: Configure your EC2 Fleet to launch instances into Capacity Blocks.
-
To set up an EKS managed node group with a Capacity Block, see Create a managed node group with Capacity Blocks for ML in the Amazon EKS User Guide.
-
To set up AWS ParallelCluster using a Capacity Block, see ML on AWS ParallelCluster
.
