SageMaker notebook instances should run on supported platforms

How to fix it

Set the notebook instance's platform identifier to notebook-al2023-v1. For an AL2 instance the move is in place and keeps the volume: stop the instance, update the identifier, start it again. Two things change with it. The package manager becomes dnf instead of yum, so a lifecycle configuration script that installs with yum needs editing first, and ml.p2, ml.p3, ml.p3dn, ml.inf1 and ml.g3 instance types are unsupported on AL2023, so an instance on one of those needs a new instance type in the same update. An AL1 instance (notebook-al1-v1) is not migrated in place and is replaced instead. Do not make the move through the Terraform route below: platform_identifier forces a new resource, so an apply destroys the notebook instance and the volume does not survive. Migrate through the console or CLI route, then set platform_identifier and run terraform apply -refresh-only so no replacement is planned.

In the console

SageMaker AI → Notebook instances → the stopped instance → Edit → Platform identifier

With the AWS CLI

aws sagemaker update-notebook-instance --notebook-instance-name <notebook-instance-name> --platform-identifier notebook-al2023-v1 --region <region>

With Terraform

resource "aws_sagemaker_notebook_instance" "this" {
  name                = var.notebook_instance_name
  instance_type       = var.instance_type
  role_arn            = var.execution_role_arn
  platform_identifier = "notebook-al2023-v1"
}