Virtual PyPI no longer finding private upstream packages

We are using the preview for Virtual PyPI repositories in Artifact Registry. Our setup is simple

  1. Private package repository
  2. Proxy to PyPI
  3. Virtual PyPI repository that points to the two previous repositories as upstream resources

This worked flawlessly until last Friday. Suddenly it can no longer find the private packages when using `pip install`. If I update the index url to point to the private repository directly, it works just fine.

Any suggestions would be appreciated!

1 5 304
5 REPLIES 5

Hi @alec_6b12,

Welcome to the Google Cloud Community!

You can try the following troubleshooting options:

  1. Make sure that you have the required IAM role in your repository. View Required Roles. To download or install packages, you must have the Artifact Registry Reader role.
  2. Make sure that you are using an updated version of the Google Cloud CLI. To update, you can use the command:
    gcloud components update
  3. Make sure that you have followed Setting up authentication to Python Package Repositories.
  4. For additional information, you can refer to this documentation about Installing Packages.
  5. If the above options don't work, you can contact Google Cloud Support to further look into your case. 

Let me know if it helped, thanks!

Thank you. These steps were all performed, and the virtual repository was working previously.

We opened up a support ticket for this very same issue and they replied with this:


We have identified the problem and working with product engineering with a fix to the problem.  The roll-out is being deployed but the ETA for when it will be resolve will be by tomorrow Tuesday, July 18th.  While that is going on, I will move the case to a P2 and provide any status updates of the completion tomorrow morning.

Haven't received an update from them yet and the issue still persists for us. Private repositories are still unreachable through a virtual Artifact Registry Repository

Thank you for passing this information on!

We received another update today - 

The roll-out for us-west1 was deployed and roll-out today.

We went ahead and tested several of our applications and the issue seems to be resolved in our region