Recent Docker BuildKit Features You're Missing Out On

With introduction of BuildKit - the improved builder backend for Docker - many new features has been added to Docker, many of which are little known. So, here's a rundown of the ones you definitely need to know about and should start using to make better use of Docker.

Debugging

Starting with the most common of tasks - debugging. Debugging docker build has always been a pain - if some RUN or COPY command fails you can hardly view context and debug what went wrong, usually resorting to adding RUN ls -la and similar to get more info. That however now changes with introduction of docker buildx debug:


export BUILDX_EXPERIMENTAL=1
docker buildx debug --invoke /bin/sh --on=error build .

[+] Building 1.2s (14/18)                docker:default
...
------
 > [builder 5/6] RUN exit 1:
------
Dockerfile:10
--------------------
   8 |     RUN pip3 install -r requirements.txt
   9 |
  10 | >>> RUN exit 1
  11 |
  12 |     COPY . /app
--------------------
ERROR: process "/bin/sh -c exit 1" did not complete successfully: exit code: 1
[+] Building 0.0s (0/0)                  docker:default
Launching interactive container. Press Ctrl-a-c to switch to monitor console
Interactive container was restarted with process "u6agxp1ywqapemxrt8iexfv4h". Press Ctrl-a-c to switch to the new container
/ # ls -la
total 72
drwxr-xr-x    1 root     root          4096 May  5 12:59 .
drwxr-xr-x    1 root     root          4096 May  5 12:59 ..
drwxr-xr-x    1 root     root          4096 May  4 10:11 app
...

As we can see in the snippet above, we start by enabling experimental BuildKit features with BUILDX_EXPERIMENTAL environment variable. We then start build through docker buildx debug - if the build fails at any point we will be dropped into the container and can explore the context and debug.

Notice that we included the --on=error option that will only start the debug session if the build fails.

For more details see debugging docs.

Environment Variables

If you've run a build with BuildKit before, then you've noticed the fancy new log output. It does look nice but it's not very practical, especially when debugging. There's however an environment variable to switch to plain log output:


export BUILDKIT_PROGRESS=plain

You can also set it to rawjson which is definitely not human-readable, but could be useful if you want to process the logs in some way.

Alternatively, if you like the TTY-based dynamic output, but dislike the colors, then you can simply change them with:


BUILDKIT_COLORS="run=green:warning=yellow:error=red:cancel=cyan" docker buildx debug --invoke /bin/sh --on=error build .

Making the output look like:

Buildx Colors

See docs for other environment variables.

Exporters

BuildKit also introduces concept of exporters, which define how the output of a build will be saved. The 2 most useful options are image and registry. image -as you could expect - saves output as a container image, while registry exporter automatically pushes to specified registry:


docker buildx build --output type=registry,name=martinheinz/testimage:latest .

All we need to do is specify --output option with type registry and the destination. This option additionally supports specifying multiple registries at once:


docker buildx build --output type=registry,\"name=docker.io/martinheinz/testimage,docker.io/martinheinz/testimage2\" .

Finally, we can also provide --cache-to and --cache-from options to e.g. use existing image from registry as a cache source:


docker buildx build --output type=registry,name=martinheinz/testimage:latest \
 --cache-to type=inline \
 --cache-from type=registry,ref=docker.io/martinheinz/testimage .

...
 => CACHED docker-image://docker.io/docker/dockerfile:1.4@sha256:9ba7531bd80fb0a8...1e24ef1a0dbc
...
 => CACHED [builder 2/5] WORKDIR /app                                                                            0.0s
 => CACHED [builder 3/5] COPY requirements.txt /app                                                              0.0s
 => CACHED [builder 4/5] RUN --mount=type=cache,target=/root/.cache/pip     pip3 install -r requirements.txt     0.0s
 => CACHED [builder 5/5] COPY . /app                                                                             0.0s
 => CACHED [dev-envs 1/3] RUN <<EOF (apk update...)                                                              0.0s
 => CACHED [dev-envs 2/3] RUN <<EOF (addgroup -S docker...)                                                      0.0s
 => CACHED [dev-envs 3/3] COPY --from=gloursdocker/docker / /                                                    0.0s
 => preparing layers for inline cache                                                                            0.0s
...

Imagetools

Simple, but handy subcommand of docker buildx called imagetools, allows us to inspect of an image in registry without having to pull it. The documentation includes a lot of examples, but the most useful one for me is getting a digest of remote image:


docker buildx imagetools inspect alpine --format "{{json .Manifest}}" | jq .digest
"sha256:c5b1261d6d3e43071626931fc004f70149baeba2c8ec672bd4f27761f8e1ad6b"

Latest Dockerfile Syntax

With BuildKit comes also new Dockerfile syntax through what's called Dockerfile frontend. To enable current latest syntax we need to add a directive to the top of the Dockerfile, e.g.:


# syntax=docker/dockerfile:1.3
FROM ...

To find version you can check dockerfile-upstream Docker Hub repository.

Here-docs

First of these Dockerfile syntax improvements I want to mention is here-docs, which allows us to pass multiline scripts into RUN and COPY commands:


# syntax = docker/dockerfile:1.3-labs
FROM debian
RUN <<eot bash
  apt-get update
  apt-get install -y vim
eot

# Same as:
RUN apt-get update && apt-get install -y vim

In the past we would have to use && if we wanted to put multiple commands into single RUN, now with here-docs, we can write a normal script.

Additionally, first line can specify interpreter, so we can - for example - write a Python script too:


# syntax = docker/dockerfile:1.3-labs
FROM python:3.6
RUN <<eot
#!/usr/bin/env python
print("hello world")
eot

COPY and ADD Features

In this new Dockerfile syntax, there are also more subtle changes and improvements to COPY and ADD in form of new options.

COPY now supports --parents option:


# syntax=docker/dockerfile:1.7.0-labs
FROM ubuntu

COPY ./one/two/some.txt /normal/

RUN find /normal
#10 [3/5] RUN find /normal
#10 0.223 /normal
#10 0.223 /normal/some.txt

COPY --parents ./one/two/some.txt /parents/

RUN find /parents
#12 [5/5] RUN find /parents
#12 0.509 /parents
#12 0.509 /parents/one
#12 0.509 /parents/one/two
#12 0.509 /parents/one/two/some.txt

If you copy a nested file with normal COPY, the image will contain only the file itself without the parent directories, with --parents the whole file tree is copied, similar to how cp --parents does it.

Similarly to --parents option, you can also use --exclude:


COPY --exclude=*.txt ./some-dir/* ./some-dest

Which will omit excluded files (and patterns) when copying.

Finally, ADD command has received an improvement too - it is now possible to directly add Git repository:


# syntax=docker/dockerfile:1.7.0-labs
FROM ubuntu

ADD git@github.com:kelseyhightower/helloworld.git /repo
RUN ls -la /repo

And when running build for this Dockerfile, we will get:


docker buildx build --ssh default --progress=plain .
#8 [2/3] ADD git@github.com:kelseyhightower/helloworld.git /repo
#8 0.478 Warning: Permanently added 'github.com' (ED25519) to the list of known hosts.
#8 1.738 ref: refs/heads/master HEAD
#8 1.738 96a652519d1aaca11085ca3a7806bead4d2c273f HEAD
#8 3.478 96a652519d1aaca11085ca3a7806bead4d2c273f refs/heads/master
#8 1.829 ref: refs/heads/master HEAD
#8 1.829 96a652519d1aaca11085ca3a7806bead4d2c273f HEAD
#8 3.838 From github.com:kelseyhightower/helloworld
#8 3.838  * [new branch]      master     -> master
#8 3.838  * [new branch]      master     -> origin/master
#8 DONE 7.4s

#9 [2/3] ADD git@github.com:kelseyhightower/helloworld.git /repo
#9 DONE 0.0s

This will also work for private repositories.

See more interesting options in docs, such ADD --keep-git-dir or ADD --checksum for validating artifact checksums

Bonus: Indentation

And while not a BuildKit feature, one thing I discovered recently is that you can indent lines in Dockerfile and it will work just fine, which allows for more readability, especially with multistage builds:


# syntax=docker/dockerfile:1
FROM golang:1.21
  WORKDIR /src

  COPY main.go .
  RUN go build -o /bin/hello ./main.go

FROM scratch
  COPY --from=0 /bin/hello /bin/hello
  CMD ["/bin/hello"]

It looks weird at first glance, but it's more readable in my opinion, making it more clear where each stage starts and which commands belong to it.

Conclusion

The examples in this article only show the features that I find the most useful, but there's plenty more, so be sure to check out official Docker docs, but also BuildKit docs, which have the latest changes. Docker blog is also a great resources, posts tagged buildkit or buildx specifically.

Subscribe: