When working with the Docker hub, two small things bothered me:
- Waiting for your build to start
- No easy control over the tags for the images.
To resolve these to issues, I created a generic Makefile that allows you to build and release docker images based upon git tags, whenever you want.
The Makefile has the following targets:
make patch-release increments the patch release level, build and push to registry
make minor-release increments the minor release level, build and push to registry
make major-release increments the major release level, build and push to registry
make release build the current release and push the image to the registry
make build builds a new version of your Docker image and tags it
make snapshot build from the current (dirty) workspace and pushes the image to the registry
make check-status will check whether there are outstanding changes
make check-release will check whether the current directory matches the tagged release in git.
make showver will show the current release tag based on the directory content.
copy the Makefile and .make-release-support into your Docker git project:
wget -O Makefile.mk https://raw.githubusercontent.com/mvanholsteijn/docker-makefile/master/Makefile
wget https://raw.githubusercontent.com/mvanholsteijn/docker-makefile/master/.make-release-support
By default, the registry is set to docker.io and the user to the current user. To override this, edit the Makefile and set the variables REGISTRY_HOST, USERNAME and NAME.
include Makefile.mk
REGISTRY_HOST=myregistry.io
USERNAME=mvanholsteijn
NAME=awesome-image
to build an image, add a Dockerfile to your directory and type make:
make
To make a release and tag it, commit add the changes and type:
make patch-release
This will bump the patch-release number, build the image and push it to the registry. It will only release if there are no outstanding changes and the content of the directory equals the tagged content.
Alternatively you can choose 'make minor-release' or 'make major-release' to bump the associated number.
The release of your docker image is kept in the file .release and uses the following format:
release=<major>.<minor>.<patch>
The name of the git tag is kept in the same file, and by default will have the format:
tag=<directory-name>.<minor>.<patch>
This will allow you to have track and tag multiple images in a single Git repository.
If you want to use a different tag prefix, change it in the .release.
The name of the image will be created as follows:
<registry-host>/<username>/<directory name>:<tag>
The tag is has the following format:
format | when |
---|---|
<release> | the contents of the directory is equal to tagged content in git |
<release>-<commit> | the contents of the directory is not equal to the tagged content |
<release>-<commit>-dirty | the contents of the directory has uncommitted changes |
If you want to maintain multiple docker images in a single git repository, you can use an alternate setup where the Makefile is located in a silbing directory.
├── multiple-example
│ ├── ...
│ ├── image1
│ │ ├── .release
│ │ ├── Dockerfile
│ │ └── Makefile
│ ├── image2
│ │ ├── .release
│ │ ├── Dockerfile
│ │ └── Makefile
│ └── make
│ ├── .make-release-support
│ ├── Makefile
The Makefile in the image directories will include the generic Makefile. In this Makefile you can alter the names and tailor the build by adding pre and post build targets. Checkout the directory (multiple-example) for an example.
To create the generic make directory, type:
mkdir make
cd make
wget https://raw.githubusercontent.com/mvanholsteijn/docker-makefile/master/Makefile
wget https://raw.githubusercontent.com/mvanholsteijn/docker-makefile/master/.make-release-support
For each docker images, you create a sibling directory:
mkdir ../image1
cd ../image1
cat > Makefile <<!
include ../make/Makefile
USERNAME=mvanholsteijn
pre-build:
@echo do some stuff before the docker build
post-build:
@echo do some stuff after the docker build
!
Now you can use the make build and release instructions to build these images.
Use the DOCKER_BUILD_CONTEXT
variable to set the path to the context. Set DOCKER_FILE_PATH
to change the path to the Dockerfile (file name included) you want to use. Using this in conjuction with the pre-build
and post-build
targets and a .dockerignore
file allows you to modify the build context.
DOCKER_BUILD_CONTEXT=../..
DOCKER_FILE_PATH=$(DOCKER_BUILD_CONTEXT)/docker/$(NAME)/some.Dockerfile
pre-build: check-env-vars .dockerignore
cp .dockerignore $(DOCKER_BUILD_CONTEXT)
post-build:
rm $(DOCKER_BUILD_CONTEXT)/.dockerignore
If you want add the current release to a source file, you can add the property pre\_tag\_command
to the .release file.
this command is executed when the .release file is updated and before the tag is placed. In the command @@RELEASE@@ is
replaced with the current release before it is executed. For example:
release=0.1.0
tag=v0.1.0
pre_tag_command=sed -i "" -e 's/^version=.*/version="@@RELEASE@@"/' setup.py
If you want to add any command line options to the docker build command, specify the Make
variable DOCKER_BUILD_ARGS
.
The following Makefile, specifies the build argument TAG_VERSION
to be set to the current value.
include ../Makefile
DOCKER_BUILD_ARGS=--build-arg TAG_VERSION=$(VERSION)
checkout the example Make- and Dockerfile.