Goal

Most of our e2e tests run against a private managed cloud. On each job run, the job will upload a new RHCOS image which will be used for testing and then deleted at the end of the job run. However, this RHCOS image for a given OCP release rarely changes and we could save significant bandwidth by only doing this when the image changes.

Why is this important?

This repeated image upload is one of the main causes for high IO in the underlying infra according to our recent investigation. We also expect to save big $$$ in bandwidth cost if we stop downloading an image for each job.

Scenarios

As a CI user, I would like to avoid repeated uploading and deletion of CI images.

Acceptance Criteria

• All OpenStack e2e tests running on the vexxhost managed cloud should be using the same image.
• (Optional) We may wish to roll this out to other clouds.

Dependencies (internal and external)

None.

Previous Work (Optional):

None.

Open questions:

• How do we avoid conflicts where two jobs run simultaneously and both decide to upload a new RHCOS image? Should we delegate refreshing of RHCOS images to a separate job, perhaps?