Feature: Delete Cluster Command

src/sagemaker/hyperpod/cli/commands/cluster_stack.py

@@ -18,6 +18,7 @@
 from sagemaker.hyperpod.common.telemetry.constants import Feature
 from sagemaker.hyperpod.common.utils import setup_logging
 from sagemaker.hyperpod.cli.utils import convert_datetimes
+from sagemaker.hyperpod import create_boto3_client
 
 logger = logging.getLogger(__name__)
 
@@ -292,8 +293,11 @@ def list_cluster_stacks(region, debug, status):
 
 @click.command("cluster-stack")
 @click.argument("stack-name", required=True)
+@click.option("--retain-resources", help="Comma-separated list of resources to retain during deletion")
+@click.option("--region", required=True, help="AWS region (required)")
 @click.option("--debug", is_flag=True, help="Enable debug logging")
-def delete(stack_name: str, debug: bool) -> None:
+@_hyperpod_telemetry_emitter(Feature.HYPERPOD_CLI, "delete_cluster_stack_cli")
+def delete_cluster_stack(stack_name: str, retain_resources: str, region: str, debug: bool) -> None:
     """Delete a HyperPod cluster stack.
 
     Removes the specified CloudFormation stack and all associated AWS resources.
@@ -305,12 +309,214 @@ def delete(stack_name: str, debug: bool) -> None:
        .. code-block:: bash
 
           # Delete a cluster stack
-          hyp delete hyp-cluster my-stack-name
+          hyp delete cluster-stack my-stack-name --region us-west-2
+
+          # Delete with retained resources (only works on DELETE_FAILED stacks)
+          hyp delete cluster-stack my-stack-name --retain-resources S3Bucket-TrainingData,EFSFileSystem-Models --region us-west-2
     """
     logger = setup_logging(logging.getLogger(__name__), debug)
 
-    logger.info(f"Deleting stack: {stack_name}")
-    logger.info("This feature is not yet implemented.")
+    try:
+        # Parse retain resources
+        retain_list = []
+        if retain_resources:
+            retain_list = [r.strip() for r in retain_resources.split(',') if r.strip()]
+
+        # Get stack resources for warning display
+        cf_client = create_boto3_client('cloudformation', region_name=region)
+
+        try:
+            resources_response = cf_client.list_stack_resources(StackName=stack_name)
+            resources = resources_response.get('StackResourceSummaries', [])
+        except Exception as e:
+            if "does not exist" in str(e):
+                click.secho(f"❌ Stack '{stack_name}' not found", fg='red')
+                return
+            raise
+
+        if not resources:
+            click.secho(f"❌ No resources found in stack '{stack_name}'", fg='red')
+            return
+
+        # Validate retain resources exist in stack
+        if retain_list:
+            existing_resource_names = {r.get('LogicalResourceId', '') for r in resources}
+            valid_retain_resources = []
+            invalid_retain_resources = []
+
+            for resource in retain_list:
+                if resource in existing_resource_names:
+                    valid_retain_resources.append(resource)
+                else:
+                    invalid_retain_resources.append(resource)
+
+            # Show warning for non-existent resources
+            if invalid_retain_resources:
+                click.secho(f"⚠️  Warning: The following {len(invalid_retain_resources)} resources don't exist in the stack:", fg='yellow')
+                for resource in invalid_retain_resources:
+                    click.secho(f" - {resource} (not found)", fg='yellow')
+                click.echo()
+
+            # Update retain_list to only include valid resources
+            retain_list = valid_retain_resources
+
+        # Categorize resources (excluding retained ones from deletion display)
+        resource_categories = {
+            'EC2 Instances': [],
+            'Networking': [],
+            'IAM': [],
+            'Storage': [],
+            'Other': []
+        }
+
+        for resource in resources:
+            resource_type = resource.get('ResourceType', '')
+            resource_name = resource.get('LogicalResourceId', '')
+            physical_id = resource.get('PhysicalResourceId', '')
+
+            # Skip resources that will be retained
+            if resource_name in retain_list:
+                continue
+
+            if 'EC2::Instance' in resource_type:
+                resource_categories['EC2 Instances'].append(f" - {resource_name} ({physical_id})")
+            elif any(net_type in resource_type for net_type in ['VPC', 'SecurityGroup', 'InternetGateway', 'Subnet', 'RouteTable']):
+                resource_categories['Networking'].append(f" - {resource_name}")
+            elif 'IAM' in resource_type:
+                resource_categories['IAM'].append(f" - {resource_name}")
+            elif any(storage_type in resource_type for storage_type in ['S3', 'EFS', 'EBS']):
+                resource_categories['Storage'].append(f" - {resource_name}")
+            else:
+                resource_categories['Other'].append(f" - {resource_name}")
+
+        # Count total resources (excluding retained ones)
+        total_resources = sum(len(category) for category in resource_categories.values())
+        retained_count = len(retain_list)
+
+        # Display warning
+        click.secho(f"⚠ WARNING: This will delete the following {total_resources} resources:", fg='yellow')
+        click.echo()
+
+        for category, items in resource_categories.items():
+            if items:
+                click.echo(f"{category} ({len(items)}):")
+                for item in items:
+                    click.echo(item)
+                click.echo()
+
+        if retain_list:
+            click.secho(f"The following {retained_count} resources will be RETAINED:", fg='green')
+            for resource in retain_list:
+                click.secho(f" ✓ {resource} (retained)", fg='green')
+            click.echo()
+
+        # Confirmation prompt
+        if not click.confirm("Continue?", default=False):
+            click.echo("Operation cancelled.")
+            return
+
+        # Perform deletion
+        delete_params = {'StackName': stack_name}
+        if retain_list:
+            delete_params['RetainResources'] = retain_list
+
+        logger.info(f"Deleting stack: {stack_name} with params: {delete_params}")
+
+        try:
+            cf_client.delete_stack(**delete_params)
+
+            click.secho(f"✓ Stack '{stack_name}' deletion initiated successfully", fg='green')
+
+            if retain_list:
+                click.echo()
+                click.secho(f"Successfully retained as requested ({len(retain_list)}):", fg='green')
+                for resource in retain_list:
+                    click.secho(f" ✓ {resource} (retained)", fg='green')
+                click.echo()
+                click.secho("💡 Retained resources will remain as standalone AWS resources", fg='cyan')
+                click.secho("   You can access them directly via AWS Console/CLI using their physical resource IDs", fg='cyan')
+
+        except Exception as delete_error:
+            # Handle termination protection specifically
+            if "TerminationProtection is enabled" in str(delete_error):
+                click.secho("❌ Stack deletion blocked: Termination Protection is enabled", fg='red')
+                click.echo()
+                click.secho("To delete this stack, first disable termination protection:", fg='yellow')
+                click.secho(f"aws cloudformation update-termination-protection --no-enable-termination-protection --stack-name {stack_name} --region {region or 'us-west-2'}", fg='cyan')
+                click.echo()
+                click.secho("Then retry the delete command.", fg='yellow')
+                raise click.ClickException("Termination protection must be disabled before deletion")
+
+            # Handle CloudFormation retain-resources limitation
+            if retain_list and "specify which resources to retain only when the stack is in the DELETE_FAILED state" in str(delete_error):
+                click.secho("❌ CloudFormation limitation: --retain-resources only works on failed deletions", fg='red')
+                click.echo()
+                click.secho("💡 Recommended workflow:", fg='yellow')
+                click.secho("1. First try deleting without --retain-resources:", fg='cyan')
+                click.secho(f"   hyp delete cluster-stack {stack_name} --region {region or 'us-west-2'}", fg='cyan')
+                click.echo()
+                click.secho("2. If deletion fails, the stack will be in DELETE_FAILED state", fg='cyan')
+                click.secho("3. Then retry with --retain-resources to keep specific resources:", fg='cyan')
+                click.secho(f"   hyp delete cluster-stack {stack_name} --retain-resources {retain_resources} --region {region or 'us-west-2'}", fg='cyan')
+                click.echo()
+                click.secho("⚠️  Alternative: Delete without retention and manually preserve important data first", fg='yellow')
+                return  # Exit gracefully without raising exception
+
+            # Handle partial deletion failures
+            click.secho("✗ Stack deletion failed", fg='red')
+
+            # Try to get current stack resources to show what was deleted
+            try:
+                current_resources = cf_client.list_stack_resources(StackName=stack_name)
+                current_resource_names = {r['LogicalResourceId'] for r in current_resources.get('StackResourceSummaries', [])}
+                original_resource_names = {r['LogicalResourceId'] for r in resources}
+
+                deleted_resources = original_resource_names - current_resource_names
+                failed_resources = current_resource_names - set(retain_list) if retain_list else current_resource_names
+
+                if deleted_resources:
+                    click.echo()
+                    click.secho(f"Successfully deleted ({len(deleted_resources)}):", fg='green')
+                    for resource in deleted_resources:
+                        click.secho(f" ✓ {resource}", fg='green')
+
+                if failed_resources:
+                    click.echo()
+                    click.secho(f"Failed to delete ({len(failed_resources)}):", fg='red')
+                    for resource in failed_resources:
+                        click.secho(f" ✗ {resource} (DependencyViolation: has dependent resources)", fg='red')
+
+                if retain_list:
+                    click.echo()
+                    click.secho(f"Successfully retained as requested ({len(retain_list)}):", fg='green')
+                    for resource in retain_list:
+                        click.secho(f" ✓ {resource} (retained)", fg='green')
+
+                click.echo()
+                click.secho("💡 Note: Some resources may have dependencies preventing deletion", fg='yellow')
+                click.secho("   Check the AWS CloudFormation console for detailed dependency information", fg='cyan')
+
+            except:
+                # If we can't get current resources, show generic error
+                click.secho(f"Error: {delete_error}", fg='red')
+
+            raise click.ClickException(str(delete_error))
+
+    except Exception as e:
+        logger.error(f"Failed to delete stack: {e}")
+        if debug:
+            logger.exception("Detailed error information:")
+
+        if "does not exist" in str(e):
+            click.secho(f"❌ Stack '{stack_name}' not found", fg='red')
+        elif "AccessDenied" in str(e):
+            click.secho("❌ Access denied. Check AWS permissions", fg='red')
+        else:
+            click.secho(f"❌ Error deleting stack: {e}", fg='red')
+
+        # Only raise exception for truly unexpected errors, not user-facing ones
+        if not any(msg in str(e) for msg in ["does not exist", "AccessDenied"]):
+            raise click.ClickException(str(e))
 
 @click.command("cluster")
 @click.option("--cluster-name", required=True, help="The name of the cluster to update")

src/sagemaker/hyperpod/cli/hyp_cli.py

@@ -10,7 +10,7 @@
 from sagemaker.hyperpod.cli.commands.cluster import list_cluster, set_cluster_context, get_cluster_context, \
     get_monitoring
 from sagemaker.hyperpod.cli.commands.cluster_stack import create_cluster_stack, describe_cluster_stack, \
-    list_cluster_stacks, update_cluster
+    list_cluster_stacks, update_cluster, delete_cluster_stack
 from sagemaker.hyperpod.cli.commands.training import (
     pytorch_create,
     list_jobs,
@@ -190,6 +190,7 @@ def exec():
 delete.add_command(pytorch_delete)
 delete.add_command(js_delete)
 delete.add_command(custom_delete)
+delete.add_command(delete_cluster_stack)
 
 list_pods.add_command(pytorch_list_pods)
 list_pods.add_command(js_list_pods)