When several goroutines call uncompress converter in parallel, the ref name
conflicts each other. This leads to Writer method failing with Unavaliable error
without retry.
For solving this issue, OpenWriter helper should be used. This allows them to
retry in such situations.
Signed-off-by: Kohei Tokunaga <ktokunaga.mail@gmail.com>
Currently, `image.GetDiffID` cannot calculate DiffID of zstd layers because it
directly uses `compress/gzip` decompressor.
This commit fixes this issue by using the generic decompressor.
Signed-off-by: Kohei Tokunaga <ktokunaga.mail@gmail.com>
Go example:
```go
opts := []converter.Opt{
// convert Docker media types to OCI ones
converter.WithDocker2OCI(true),
// convert tar.gz layers to uncompressed tar layers
converter.WithLayerConvertFunc(uncompress.LayerConvertFunc),
}
srcRef := "example.com/foo:orig"
dstRef := "example.com/foo:converted"
dstImg, err = converter.Convert(ctx, client, dstRef, srcRef, opts...)
fmt.Println(dstImg.Target)
```
ctr example: `ctr images convert --oci --uncompress example.com/foo:orig example.com/foo:converted`
Go test: `go test -exec sudo -test.root -test.run TestConvert`
The implementation is from https://github.com/containerd/stargz-snapshotter/pull/224,
but eStargz-specific functions are not included in this PR.
eStargz converter can be specified by importing `estargz` package and using `WithLayerConvertFunc(estargz.LayerConvertFunc)` option.
This converter interface will be potentially useful for converting zstd and ocicrypt layers as well.
Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
zstd is a compression algorithm that has a very fast decoder, while
providing also good compression ratios. The fast decoder makes it
suitable for container images, as decompressing the tarballs is a very
expensive operation.
https://github.com/opencontainers/image-spec/pull/788 added support
for zstd to the OCI image specs.
Signed-off-by: Giuseppe Scrivano <gscrivan@redhat.com>
Give control of the content labeling process for children to
the client. This allows the client to control the names
associated with the labels and filter out labels.
Signed-off-by: Derek McGowan <derek@mcg.dev>
This log message shows up in the client's logs. For any media type that
the client doesn't know about it will wind up with a warning log.
Downgrade this to debug since it is more of a development concern.
We encountered this trying to fetch Docker plugins which has a media
type for plugin configs.
Signed-off-by: Brian Goff <cpuguy83@gmail.com>
Dependencies may be switching to use the new `%w` formatting
option to wrap errors; switching to use `errors.Is()` makes
sure that we are still able to unwrap the error and detect the
underlying cause.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Avoid directly handling media types with "+" attributes,
instead handling the base and passing through the full
media type to the appropriate stream processor or decompression.
Signed-off-by: Derek McGowan <derek@mcgstyle.net>
Prevent an errgroup error from causing the acquire to
return a cancellation error. Previously any error
from the errgroup would cause the Dispatch to always
return the cancelled error.
Signed-off-by: Derek McGowan <derek@mcgstyle.net>
We are separating out the encryption code and have designed a few new
interfaces and APIs for processing content streams. This keep the core
clean of encryption code but enables not only encryption but support of
multiple content types ( custom media types ).
Signed-off-by: Michael Crosby <crosbymichael@gmail.com>
client.Pull will only pull one matching platform by default.
When checking the size of image we match that behavior so that
we don't look for multiple platforms that might not exist on disk.
Signed-off-by: Darren Shepherd <darren@rancher.com>
When providing multiple names, the shared annotation map was
causing the names to get overridden.
Combined the WithManifest options which had compatible interfaces.
Signed-off-by: Derek McGowan <derek@mcgstyle.net>
When client uses Pull action to pull image, it will limit the number of
manifest as one. But Unpack action will call Manifest to traverse all
the manifests including non-dowloaded one. If the platform has more than
one manifest, the Pull with unpack action will fail. And also, there is
no need to read non-best matched manifest. Therefore, the Manifest can
do the sort earlier.
Signed-off-by: Wei Fu <fuweid89@gmail.com>
Remove annotation prefix and add multiple index records
for manifests with multiple image names. This makes the
custom annotation more consistent with the OCI image
annotation. Additionally, ensure the OCI image annotation
always represents the tag (partial image name) as recommended
by the specification. The containerd image name annotation
will always contain the full image name.
Signed-off-by: Derek McGowan <derek@mcgstyle.net>
Ensure the manifest which gets created using the Docker
compatibility code compresses the blob before creating
the manifest. This ensures consistency with manifests
used by Docker.
Signed-off-by: Derek McGowan <derek@mcgstyle.net>
Using the cri utility caused other project to have
containerd/cri as a dependency, only for this utility.
The new `reference.ParseDockerRef` function does the
same (other than having a different name).
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
