docker: extracting a layer from a image

In this specific case, it looks like the ADD command added the base image to the file system. If you run docker history --no-trunc docker/whalesay, the full command is:

/bin/sh -c #(nop) ADD file:f4d7b4b3402b5c53f266bb7fdd7e728493d9a17f9ef20c8cb1b4759b6e66b70f in / 

docker history reports that particular layer is 188MB. Let's look at these layers in more detail:

$ docker save docker/whalesay -o whalesay.tar $ tar tvf whalesay.tar ... -rw-r--r-- 0/0 197181952 2015-05-25 22:04 cc88f763e297503d2407d6b462b2b390a6fd006b30f51c8efa03dd88fa801b89/layer.tar ... 

Looks like a pretty good candidate! You can now extract that layer and pull files out of it.

$ tar xf whalesay.tar cc88f763e297503d2407d6b462b2b390a6fd006b30f51c8efa03dd88fa801b89/layer.tar $ tar xf cc88f763e297503d2407d6b462b2b390a6fd006b30f51c8efa03dd88fa801b89/layer.tar etc/passwd 

If you're looking to pull a particular file out of a layer, but you don't know which layer, you could do this. First, extract all the layers:

$ tar xf whalesay.tar 

Now you've got all the layers as individual .tar files. Let's find a file:

$ for layer in */layer.tar; do tar -tf $layer | grep docker.cow && echo $layer; done usr/local/share/cows/docker.cow 0523c5a0c4588dde33d61d171c41c2dc5c829db359f4d56ab896ab1c185ed936/layer.tar cowsay/cows/docker.cow 40e8ae7bb4e5b9eaac56f5be7aa614ed50f163020c87ba59e905e01ef0af0a4f/layer.tar cowsay/cows/docker.cow f9bc8676543761ff3033813257937aeb77e9bc84296eaf025e27fe01643927cf/layer.tar 

Finally, extract the file from the layer you want:

$ tar xf 0523c5a0c4588dde33d61d171c41c2dc5c829db359f4d56ab896ab1c185ed936/layer.tar \ usr/local/share/cows/docker.cow 

This will extract that file with the full path relative to the current directory.

$ cat usr/local/share/cows/docker.cow ## ## Docker Cow ## $the_cow = <<EOC; $thoughts $thoughts $thoughts ## . ## ## ## == ## ## ## ## === /""""""""""""""""\___/ === ~~~ {~~ ~~~~ ~~~ ~~~~ ~~ ~ / ===- ~~~ \\______ o __/ \\ \\ __/ \\____\\______/ EOC 

It looks like other people would also like to have this feature, but unfortunately right now it does not seem to exist.

See also this issue and here a related request which got rejected.

If you're fine with saving the complete docker (docker save) and then extracting a tarball with your layer, then this is possible:

docker run -it <your image> # do fancy stuff in the container docker commit <your container> foobar # create image from container docker history foobar # will show you the layers docker save -o foobar.tar foobar # dumps container contents to foobar.tar 

Now foobar.tar will contain the file system states from different times. Inspecting this tarball shows, in my case, a file repositories with

{"foobar":{"latest":"fdf43d96e691c57e9afb4b85dba2e6745146a7ca9076c7284c6b2e1f93434562"}} 

which indicates, that the latest layer is fdf43.... You can get a tarball with the file system contents of this layer via

tar -x fdf43d96e691c57e9afb4b85dba2e6745146a7ca9076c7284c6b2e1f93434562/layer.tar -f foobar.tar 

There is a tool, undocker, which automated this process, but I'm not sure whether it will work with the current format of the saved tar file.

I don't really understand what you mean by "extract" but if you want to get further information about image, run

docker inspect <image_name> 

You you want to get file, then run container from this image. Try

docker export <container_name> > abc.tar 

after that, extract abc.tar and find your file.

While not capable of extracting a specific layer, the docker-save-last-layer command line utility is made to extract the last layer only. Combined with docker build --squash you can avoid exporting the base layers. This may help to accomplish your goals.

It works by using a patched version of the docker daemon inside a docker image that can access the images on your host machine. So it doesn't require doing a full docker save before using it. This makes it performant for large base images.

Typical usage is simple and looks like:

pip install d-save-last docker build --t myimage --squash . d-save-last myimage -o ./myimage.tar 

Docker is not capable of saving layers individually, however there is a tool on Github called dlgrab that claims to do it. https://github.com/aidanhs/dlgrab

The regclient/regclient project does include a regctl blob/regctl later which allows to fetch a specific layer.

You can list layers with regctl image manifest.

Since regctl v0.4.5, you can even use (PR 296):

• regctl blob get-file to fetch a file from a layer
• regctl image get-file to fetch a file from the image layers

Example:

$ regctl image manifest busybox { "schemaVersion": 2, "mediaType": "application/vnd.docker.distribution.manifest.v2+json", "config": { "mediaType": "application/vnd.docker.container.image.v1+json", "size": 1493, "digest": "sha256:6858809bf669cc5da7cb6af83d0fae838284d12e1be0182f92f6bd96559873e3" }, "layers": [ { "mediaType": "application/vnd.docker.image.rootfs.diff.tar.gzip", "size": 764618, "digest": "sha256:df8698476c65c2ee7ca0e9dbc2b1c8b1c91bce555819a9aaab724ac64241ba67" } ] } $ regctl blob get busybox sha256:6858809bf669cc5da7cb6af83d0fae838284d12e1be0182f92f6bd96559873e3 | jq . { "architecture": "amd64", "config": { "Hostname": "", "Domainname": "", "User": "", "AttachStdin": false, "AttachStdout": false, "AttachStderr": false, "Tty": false, "OpenStdin": false, "StdinOnce": false, ... 

For docker images that contain one layer.tar, this works.

docker save myimage:latest | tar xO --wildcards '*.tar' | tar xv myfile 

First step extracts myimage as tar to stdout. In the second step the '*.tar', which is in a dockerfile is usually layer.tar is selected and untared to stdout . In the third step we get the layer.tar from stdout and untar whatever file we want to pick