- Newest
- Most votes
- Most comments
I understand your concern about unexpected SageMaker charges and the need to ensure all resources are properly terminated. Let me provide some guidance on how to address this situation and prevent future charges.
First, it's important to note that simply deleting a notebook instance in SageMaker does not automatically terminate all associated resources. To fully clean up SageMaker resources and avoid unnecessary charges, you need to follow a comprehensive process:
- Delete the endpoint (if any). This also removes the associated ML compute instances.
- Delete the endpoint configuration.
- Delete the model.
- Stop and then delete the notebook instance.
- Delete the S3 bucket used for storing model artifacts and training datasets.
- Delete CloudWatch log groups with names starting with "/aws/sagemaker/".
In your case, it appears that some resources, particularly an instance (mlt3.medium) and a storage volume, remained active after your initial deletion attempt. This explains the charges you've seen.
To ensure all resources are now terminated:
- Open the SageMaker console and check for any running instances, endpoints, or models.
- In the S3 console, look for any remaining SageMaker-related buckets.
- In CloudWatch, search for and delete any SageMaker-related log groups.
Regarding your refund request, while I can't guarantee a refund, AWS does have a history of being understanding in cases of accidental usage. I recommend contacting AWS Support directly to explain the situation and request a one-time refund for these unintended charges.
For future reference, always double-check that all components are shut down when you're done using SageMaker. Remember that shutting down a notebook running on an instance is not the same as shutting down the instance itself. You must explicitly stop or terminate instances to avoid ongoing charges.
By following these steps, you should be able to confirm that all SageMaker resources are fully deactivated, preventing any further unexpected charges.
Sources
Clean up - Amazon SageMaker
Usage Metering - Amazon SageMaker
Clean up Amazon SageMaker notebook instance resources - Amazon SageMaker
Relevant content
- asked 3 months ago
- AWS OFFICIALUpdated 6 months ago
- AWS OFFICIALUpdated 6 months ago
