batchReplaceClusterNodes method

Future<BatchReplaceClusterNodesResponse> batchReplaceClusterNodes({ required String clusterName, List<String>? nodeIds, List<String>? nodeLogicalIds, })

Replaces specific nodes within a SageMaker HyperPod cluster with new hardware. BatchReplaceClusterNodes terminates the specified instances and provisions new replacement instances with the same configuration but fresh hardware. The Amazon Machine Image (AMI) and instance configuration remain the same.

This operation is useful for recovering from hardware failures or persistent issues that cannot be resolved through a reboot.

• Data Loss Warning: Replacing nodes destroys all instance volumes, including both root and secondary volumes. All data stored on these volumes will be permanently lost and cannot be recovered.
• To safeguard your work, back up your data to Amazon S3 or an FSx for Lustre file system before invoking the API on a worker node group. This will help prevent any potential data loss from the instance root volume. For more information about backup, see Use the backup script provided by SageMaker HyperPod.
• If you want to invoke this API on an existing cluster, you'll first need to patch the cluster by running the UpdateClusterSoftware API. For more information about patching a cluster, see Update the SageMaker HyperPod platform software of a cluster.
• You can replace up to 25 nodes in a single request.

May throw ResourceNotFound.

Parameter clusterName : The name or Amazon Resource Name (ARN) of the SageMaker HyperPod cluster containing the nodes to replace.

Parameter nodeIds : A list of EC2 instance IDs to replace with new hardware. You can specify between 1 and 25 instance IDs.

• Either NodeIds or NodeLogicalIds must be provided (or both), but at least one is required.
• Each instance ID must follow the pattern i- followed by 17 hexadecimal characters (for example, i-0123456789abcdef0).
• For SageMaker HyperPod clusters using the Slurm workload manager, you cannot replace instances that are configured as Slurm controller nodes.

Parameter nodeLogicalIds : A list of logical node IDs to replace with new hardware. You can specify between 1 and 25 logical node IDs.

The NodeLogicalId is a unique identifier that persists throughout the node's lifecycle and can be used to track nodes that are still being provisioned and don't yet have an EC2 instance ID assigned.

• Replace operations destroy all instance volumes (root and secondary). Ensure you have backed up any important data before proceeding.
• This parameter is only supported for clusters using Continuous as the NodeProvisioningMode. For clusters using the default provisioning mode, use NodeIds instead.
• Either NodeIds or NodeLogicalIds must be provided (or both), but at least one is required.

Implementation

Future<BatchReplaceClusterNodesResponse> batchReplaceClusterNodes({
  required String clusterName,
  List<String>? nodeIds,
  List<String>? nodeLogicalIds,
}) async {
  final headers = <String, String>{
    'Content-Type': 'application/x-amz-json-1.1',
    'X-Amz-Target': 'SageMaker.BatchReplaceClusterNodes'
  };
  final jsonResponse = await _protocol.send(
    method: 'POST',
    requestUri: '/',
    exceptionFnMap: _exceptionFns,
    // TODO queryParams
    headers: headers,
    payload: {
      'ClusterName': clusterName,
      if (nodeIds != null) 'NodeIds': nodeIds,
      if (nodeLogicalIds != null) 'NodeLogicalIds': nodeLogicalIds,
    },
  );

  return BatchReplaceClusterNodesResponse.fromJson(jsonResponse.body);
}