This issue was reported by a Red Hat partner, who observed a significant increase (~2.2 GB) in the size of the OS ISO generated using the new image preload workflow introduced to fix container image upgrade issuesRHEL-75827

Old build workflow:

1. Used podman pull to embed container images directly into the image
2. No duplication of image layers

New build workflow:

#Uses skopeo copy --all to preload images into /usr/lib/containers/storage/<sha>

A runtime script copies the images again from this path into containers-storage: (i.e., /var/lib/containers/storage), via:

skopeo copy --preserve-digests \
"dir:${IMAGE_STORAGE_DIR}/\${sha}" \
"containers-storage:\${img}"

1. Problem:

This results in each image being stored twice:

Once in the preloaded image store baked into the ISO (/usr/lib/containers/storage)

Again in the runtime container store (/var/lib/containers/storage)

Ask:
Can we optimize the image preload method in MicroShift.