Trying to update PATH or PYTHONPATH env vars for a PyTorchModel endpoint results in UnexpectedStatusException

[setu4993]

Describe the bug
Creating a PyTorchModel-based endpoint that attempts to update the PATH or PYTHONPATH environment variables results in a UnexpectedStatusException.

To reproduce

Create any PyTorchModel object and add: env={"PATH": "/opt/ml/model/code/lib:${PATH}"} or env={"PYTHONPATH": "/opt/ml/model/code/lib:${PYTHONPATH}"} during model definition.

Expected behavior

The model should be created and endpoint deployed successfully.

Screenshots or logs
Traceback:

Error traceback:
------------------
UnexpectedStatusException                 Traceback (most recent call last)
<ipython-input-15-a29e7afc3eb6> in <module>
----> 1 pt_model.deploy(initial_instance_count=1, instance_type="ml.t2.medium", endpoint_name="coach-notes-test-2020-08-25")

/opt/conda/lib/python3.6/site-packages/sagemaker/model.py in deploy(self, initial_instance_count, instance_type, serializer, deserializer, accelerator_type, endpoint_name, tags, kms_key, wait, data_capture_config)
    508             kms_key=kms_key,
    509             wait=wait,
--> 510             data_capture_config_dict=data_capture_config_dict,
    511         )
    512 

/opt/conda/lib/python3.6/site-packages/sagemaker/session.py in endpoint_from_production_variants(self, name, production_variants, tags, kms_key, wait, data_capture_config_dict)
   3095 
   3096             self.sagemaker_client.create_endpoint_config(**config_options)
-> 3097         return self.create_endpoint(endpoint_name=name, config_name=name, tags=tags, wait=wait)
   3098 
   3099     def expand_role(self, role):

/opt/conda/lib/python3.6/site-packages/sagemaker/session.py in create_endpoint(self, endpoint_name, config_name, tags, wait)
   2615         )
   2616         if wait:
-> 2617             self.wait_for_endpoint(endpoint_name)
   2618         return endpoint_name
   2619 

/opt/conda/lib/python3.6/site-packages/sagemaker/session.py in wait_for_endpoint(self, endpoint, poll)
   2884                 ),
   2885                 allowed_statuses=["InService"],
-> 2886                 actual_status=status,
   2887             )
   2888         return desc

UnexpectedStatusException: Error hosting endpoint coach-notes-test-2020-08-25: Failed. Reason: Request to service failed. If failure persists after retry, contact customer support..

System information
A description of your system. Please provide:

• SageMaker Python SDK version: 2.5.0
• Framework name (eg. PyTorch) or algorithm (eg. KMeans): PyTorch
• Framework version: 1.5.0
• Python version: 3.6
• CPU or GPU: CPU
• Custom Docker image (Y/N): No

Additional context

Updating PATH is required because of the bug mentioned in #1832.

[setu4993]

I was able to resolve this by setting PYTHONPATH without trying to use the existing env var PYTHONPATH. So, the env var definition would be: env={"PYTHONPATH": "/opt/ml/model/code/lib"}.

And then the container build step takes care of adding in the other env vars that are required. The final PYTHONPATH in the container, then is: '/.sagemaker/mms/models/model/code:sagemaker_pytorch_serving_container.handler_service:/opt/ml/model/code:/opt/ml/model/code/lib:/.sagemaker/mms/models/model'.

Ideally, this shouldn't have been breaking but I'm fine with closing this.