Skip to content

feat: support nydus image format#566

Open
fatelei wants to merge 2 commits intocontainers:mainfrom
fatelei:issue-565
Open

feat: support nydus image format#566
fatelei wants to merge 2 commits intocontainers:mainfrom
fatelei:issue-565

Conversation

@fatelei
Copy link

@fatelei fatelei commented Dec 23, 2025

resolve #565

support nydus image format

@github-actions github-actions bot added the image Related to "image" package label Dec 23, 2025
@packit-as-a-service
Copy link

Packit jobs failed. @containers/packit-build please check.

podmanbot pushed a commit to podmanbot/buildah that referenced this pull request Dec 23, 2025
@podmanbot
Copy link

✅ A new PR has been created in buildah to vendor these changes: containers/buildah#6610

Copy link
Contributor

@mtrmac mtrmac left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks!

Per https://github.com/containers/container-libs/blob/main/CONTRIBUTING.md#sign-your-prs :

Use a real name (sorry, no anonymous contributions)

(please read on for more nuance to that.)


Cc: @giuseppe (Referencing https://github.com/dragonflyoss/nydus/blob/master/docs/nydus-design.md, that says:

  • The image format has an artifactType in the index, but not in the manifest itself?!
  • It is not really an artifact, it has the usual OCI config type
  • … it’s “just” a new structure of the layer blobs.

Reading this PR, I don’t think much of these changes is actually necessary. Do copy operations fail otherwise??


Apart from pure unchanged copies, long-term we need to have some idea of what this format is from the end-user workflow perspective, and I think it’s necessary to think about that now, to shape the API / UI appropriately.

Is it a compression format, like zstd:chunked, where each individual layer can be ~losslessly converted between tar and the other compression? (Seems not to be the case, with the bootstrap/blob differentiation.)

Is it a ~different manifest format, really distinct from OCI? (Looking at the metadata format suggests “no”; in particular, returning NonImageArtifactError from OCIConfig seems inappropriate.)

And yet, it is ultimately significantly different, at least in that, right now, podman pull would fail.

How would a conversion from a tar format to Nydus, or back, work? Skopeo is (well, currently) not set up to do conversions that need to download and significantly manipulate large blobs — conversions between v2s1/v2s2/OCI are metadata operations + compression/decompression where we can stream the data with little local storage. Would these conversions ~forever be delegated to specialized tools? Or is the idea that podman build+podman push should be able to somehow directly convert to Nydus on the fly?

imgspecv1.MediaTypeLayoutHeader,
ociencspec.MediaTypeLayerEnc, ociencspec.MediaTypeLayerGzipEnc:
ociencspec.MediaTypeLayerEnc, ociencspec.MediaTypeLayerGzipEnc,
manifest.NydusBootstrapLayerMediaType, manifest.NydusBlobLayerMediaType:
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This function is deprecated and has no known users (at least within the Podman+friends tools), so I don’t think there is any point in changing it.

Signed-off-by: fatelei <fatelei@gmail.com>
Signed-off-by: fatelei <fatelei@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

image Related to "image" package

Projects

None yet

Development

Successfully merging this pull request may close these issues.

support nydus image format

3 participants