How to capture the sagemaker error in case it fails and notify via SES,SNS

Question

I have a lambda function that creates a sagemaker processing job. Now let's say the sagemaker job fails due an algorithm error or an API error etc, How do I capture the exact error message(for ex, see picture) and send an email from the same lambda function or it can be a separate event?

https://anonfiles.com/d308Jf15ue/2021-06-17_22_36_21-Amazon_SageMaker_png

Answer 1

Here's what I did. I used cloudwatch events for monitoring and in event rule target I gave SNS topic to which my email which was subscribed. Here's the event Pattern that i used.

{
  "source": ["aws.sagemaker"],
  "detail-type": ["SageMaker Processing Job State Change"],
  "detail": {
    "ProcessingJobStatus": ["Failed"]
  }
}

Cloud watch event target also has input transformer where you can fetch data from received cloudtrail event and pass it to SNS.The event data should have the error message.

https://docs.aws.amazon.com/AmazonCloudWatch/latest/events/CloudWatch-Events-Input-Transformer-Tutorial.html