training and inference run scripts support STS endpoints.

[verdimrc]

🐛 Bug report

• [x] I have checked that this issue has not already been reported.

Describe the bug

Without public internet, but with STS endpoint setup, both inference_pipeline_run.py and training_pipeline_run.py will fail due to http timeout.

This can be fix by hardcoding the endpoint as follows:

inference_pipeline_run.py:136:    sts = boto3.client("sts", endpoint_url="https://sts.ap-southeast-2.amazonaws.com")
training_pipeline_run.py:132:    sts = boto3.client("sts", endpoint_url="https://sts.ap-southeast-2.amazonaws.com")

As a proper fix, I propose to add a new configurable parameter to define the VPC endpoint for STS.

To reproduce

Run {training,inference}_pipeline_run.py scripts from an EC2 instance running in private VPC with STS endpoint.

Expected behavior

Training or inference should complete.

System information

• awscli version: aws-cli/1.18.179 Python/3.6.10 Linux/4.14.203-156.332.amzn2.x86_64 botocore/1.20.30
• SageMaker Python SDK version: 1.72.1
• Docker image: N/A
• Python version: Python/3.6.10 Linux/4.14.203-156.332.amzn2.x86_64

[github-actions[bot]]

This issue is stale. If left untouched, it will be automatically closed in 7 days.

[github-actions[bot]]

This issue is stale because it has been open for 60 days with no activity. Please update or respond to this comment if you're still interested in working on this.