For https://fedoraproject.org/wiki/Changes/zstd:chunked (rejected for F41 but we would like to submit it again for a future release, ideally F42), we would need cloud-image-uploader to learn to generate manifests with both gzip & zstd:chunked images.
Note that compressions happens when pushing for container images thus this is the place where this should happen.
The flow would be: - Upload the image for all architectures, using 'gzip' compression - Repeat using 'zstd:chunked' compression - Build a manifest out of all of those images
@dwalsh FYI
Seems pretty straightforward to add.
Regarding rolling it out, does it need to only happen for a particular version, or is it transparent to clients that aren't aware of additional compression algorithms?
I'm sure the answer is somewhere in the associated discussions thread, but I admit I only skimmed it.
it is transparent to clients, the zstd:chunked version will be used only by clients that know how to use it
Log in to comment on this ticket.