Shan Dou
Shan Dou

Reputation: 3278

Why is AWS Sagemaker notebook instance designed to only persist data under ~/Sagemaker?

In my current job we use AWS managed notebooks on Sagemaker EC2. I am largely okay with the user experience but the lack of data persistency outside ~/Sagemaker has been quite inconvenient. Every time should the instance need restarting, I'd lose all the settings and python packages. Wonder why AWS would make this particular decision for Sagemaker. Have used Google Cloud's AI platform before and it does not have such settings and my configurations would always persist.

Upvotes: 1

Views: 1186

Answers (1)

rok
rok

Reputation: 2775

I faced a similar issue on other AWS services. Usually for managed services AWS uses read-only containers approach and leave just one folder of the filesystem for read/write that persist across the stop/restart cycle. Reguarding the packages installation the seems to be to install your custom environment on the notebook instance's Amazon EBS volume, as described here

Upvotes: 2

Related Questions