From Gentoo Wiki
(Redirected from Kangie/Kubler)
Jump to:navigation Jump to:search

Kubler is a generic, extendable build orchestrator, written in Bash. It can be used to take advantage of Portage's features to build lightweight Docker or Podman images without needing to mess with crossdev, or as a tool to assist with ebuild development.


Kubler may be installed from the kubler ebuild repository:

root #eselect repository add kubler git https://github.com/edannenberg/kubler-overlay.git
root #emerge --sync
root #emerge --ask kubler

Building images requires a working Docker or Podman installation.

Useful Notes

Environmental Configuration

Due to containerisation environmental configurations will not make it through to builder containers. This is often desirable as it enforces the documentation (through build-time configuration) of any non-standard configuration in the environment. To work around this many variables may be set in kubler.conf for either the namespace (global settings) or an individual image.

FILE kubler.conf

Fixing Broken Builders

Eventually the installed version of Kubler and the Gentoo repository will come out of sync; typically the first sign of this is that Kubler becomes unable to create updated Portage containers. When this happens:

root #emerge @live-rebuild
user $kubler clean
user $kubler build -C larry

Image Creation

Kubler provides a set of commands to create new images, builders and namespaces. These commands should be run from within a project directory, e.g.: /data/development/gentoo-containers/

user $mkdir -p /data/development/gentoo-containers/
user $cd /data/development/gentoo-containers/

Create a new namespace

This example will create a new namespace called 'larry' which may contain multiple images. The namespace will be created in the project directory.

Use the new command to take care of the boilerplate; choose 'multi' when asked for the namespace type:

user $kubler new namespace larry
»»» <enter> to accept default value
»»» Working dir type? Choices:
»»»   single - You can't add further namespaces to the created working dir, it only holds images
»»»   multi  - Creates a working dir that can hold multiple namespaces
»[?]» Type (single): multi
»»» Top level directory name for new namespace 'larry'? The directory is created at /data/development/gentoo-containers/
»[?]» Namespaces Dir (kubler-images):
»»»»» Initial image tag, a.k.a. version?
»[?]» Image Tag (20230706):
»[!]» New namespace location:  /data/development/gentoo-containers/kubler-images/larry
»»»»» Who maintains the new namespace?
»[?]» Name (Your Name): Larry the Cow
»[?]» EMail (your@mail.org): Larry.the.Cow@gentoo.zip
»»»»» Default build engine?
»[?]» Engine (docker):
»[✔]» Successfully created "larry" namespace at /data/development/gentoo-containers/kubler-images
»[!]» Configuration file: /data/development/gentoo-containers/kubler-images/larry/kubler.conf
»[!]» To manage the new namespace with GIT you may want to run:
»»» $ git init /data/development/gentoo-containers/kubler-images/larry
»[!]» To create images in the new namespace run:
»»» $ cd /data/development/gentoo-containers/kubler-images/larry
    $ kubler new image larry/<image_name>
user $cd larry/

Although not strictly required, installing Kubler's example images is a good idea.

user $kubler update

It is worthwhile to begin tracking this new namespace with Git. Kubler will prepopulate a .gitignore file for convenience.

user $pushd /data/development/gentoo-containers/kubler-images/larry
user $git init .
user $git add .
user $git commit -m "Initial commit"
user $popd

Create a new image

Kubler may be used to create container images for a variety of purposes. This example will create a new image called 'openldap' within the existing 'larry' namespace, based on the 'kubler/busybox' image.

user $kubler new image larry/openldap
»»» <enter> to accept default value
»»» Extend an existing Kubler managed image? Fully qualified image id (i.e. kubler/busybox) or scratch
»[?]» Parent Image (scratch): kubler/busybox
»»» Add test template(s)? Possible choices:
»»»   hc  - Add a stub for Docker's HEALTH-CHECK, recommended for images that run daemons
»»»   bt  - Add a stub for a custom build-test.sh script, a good choice if HEALTH-CHECK is not suitable
»»»   yes - Add stubs for both test types
»»»   no  - Fck it, we'll do it live!
»[?]» Tests (hc): bt
»[✔]» Successfully created new image at /data/development/gentoo-containers/kubler-images/larry/images/openldap

Use interactive build to plan and test

This step is not required; it is possible to directly edit the config/build.sh file if you are familiar with Portage.

Kubler brings a unique feature to the table when constructing an container image: the --interactive build argument. As the name implies, this launches the build container in an interactive manner, enabling users to investigate the current / inherited configuration.

user $kubler build larry/openldap -i

This will build any missing parent images/builders; the first run may take quite a bit of time - once the local binary package cache and build containers are seeded future runs will be much faster. Once the prerequisite images are ready the build container will present a shell.

For first-time users it may be convenient to search for the openldap package to ensure that the correct atom is selected and investigate any USE flags that are of interest:

root #eix openldap
* net-nds/openldap
     Available versions:  2.4.59-r2^t 2.5.14(0/2.5)^t 2.6.3-r7(0/2.6)^t ~2.6.4-r1(0/2.6)^t ~2.6.4-r2(0/2.6)^t {argon2 autoca +berkdb +cleartext crypt cxx debug experimental gnutls iodbc ipv6 kerberos kinit minimal odbc overlays pbkdf2 perl samba sasl selinux sha2 smbkrb5passwd ssl static-libs +syslog systemd tcpd test ABI_MIPS="n32 n64 o32" ABI_S390="32 64" ABI_X86="32 64 x32"}
     Homepage:            https://www.openldap.org/
     Description:         LDAP suite of application and development tools

Edit the image's build script:

root #nano /config/build.sh
The /config directory in the build container is the host mounted image directory at larry/images/openldap/. Feel free to use a local IDE/editor to edit build.sh instead.

Add the net-nds/openldap and net-misc/curl packages to the _packages variable in build.sh, update cURL USE flags, enable the ~arch versions of the packages we care about:

FILE /config/build.sh
_packages="net-nds/openldap net-misc/curl"
    # Update a Gentoo package use flag.
    update_use 'net-misc/curl' '+ldap'
    # ..or a Gentoo package keyword
    update_keywords 'net-misc/curl' '+~amd64'
    update_keywords 'net-nds/openldap' '+~amd64'
If using the busybox image as a parent, unset the su USE flag from sys-apps/util-linux in the config/build.sh file.

Perform a test run of the first build phase:

root #kubler-build-root

Once this completes successfully exit the interactive builder using exit.

Build the image

Assuming that build.sh has been configured as described above, it should be safe to attempt to build the image.

user $kubler build larry/openldap -nF
»[✘]»[larry/openldap]» fatal: build-test.sh for image larry/openldap:20230704 failed with exit signal: 1
The arguments are short hand for --no-deps and --force-full-image-build, omitting -n would also rebuild all parent images, which is waste of time in this case.

The build will fail, as expected, due to the build-test.sh script not being implemented. This is a good time to implement the build-test.sh script, which will be used to verify that the image is functional.

pipefail will cause build-test.sh to fail on busybox-based images
FILE build-test.sh
#!/usr/bin/env sh

set -eo

# Do some tests and exit with either 0 for healthy or 1 for unhealthy
# Check that the openldap bin launches and provides some expected output
/usr/lib/openldap/lloadd -VV  2>&1 | grep "OpenLDAP" || exit 1

exit 0

Unfortunately this image is not suitable for a build-time docker health check via the docker-healthcheck.sh mechanism, so it will not be demonstrated here.

Modify the image's Dockerfile.template to add any finishing touches, such as the ENTRYPOINT or CMD directives. In this example the container will act as an LDAP proxy via lloadd; additional configuration will be provided at runtime.

FILE Dockerfile.template
LABEL maintainer="${MAINTAINER}"

ADD rootfs.tar /

#COPY docker-healthcheck.sh /usr/bin/docker-healthcheck
#HEALTHCHECK --interval=60s --timeout=5s --start-period=5s --retries=3 CMD ["docker-healthcheck"]

CMD ["/usr/lib/openldap/lloadd"]

Re-run the build:

user $kubler build larry/openldap -nF
»[✔]»[larry/openldap]» done.

At this point the image should exist in the local Docker/Podman registry and be ready for use:

user $docker images
REPOSITORY                                              TAG                       IMAGE ID       CREATED          SIZE
larry/openldap                                          20230704                  09347c55282b   2 minutes ago    56.4MB
larry/openldap                                          latest                    09347c55282b   2 minutes ago    56.4MB

Pushing Images to a Docker Repository

To push images to Docker Hub:

user $kubler push larry/openldap larry/openldap

The default assumes that the given namespace equals the respective Docker Hub account names, i.e. openldap and larry. To override this you may place a push.conf file in each namespace dir with the following format:

FILE push.conf

Ebuild Development with Kubler

Kubler may be used to develop against any ebuild repository. Create a new namespace, let's call it edev:

user $kubler new namespace edev

Create a new builder, select kubler/bob as IMAGE_PARENT:

user $kubler new builder edev/bob

Edit the new builder's build.sh and add any additional repositories:

For ::gentoo only there's no need to add the repo, it's already available; binding in the development fork and updating .bashrc is probably desirable, however.
FILE /config/build.sh
configure_builder() {
    # we overwrite this with a local host mount later, but this takes care of the initial overlay setup in the builder for us
    add_overlay kubler https://github.com/edannenberg/kubler-overlay.git
    # just for convenience
    echo 'cd /var/db/repos/kubler' >> ~/.bashrc

Create a new image, let's call it bench, use kubler/bash as IMAGE_PARENT:

user $kubler new image edev/bench

Edit the new image's build.conf and configure the builder and ebuild repo to be mounted builder:

FILE /config/build.conf

Start an interactive build container and get tinkering:

user $kubler build -i edev/bench
root #ebuild dev-lang/foo/foo-0.4.0.ebuild manifest merge

Updating Build Containers

Portage is updated frequently and the cached ebuild repository will quickly become stale. To check for new releases:

user $kubler update

This will also check for updates to the example images provided by Kubler, usually updated at the end of each month. If updates were found found simply rebuild the stack by running:

user $kubler clean
user $kubler build -C namespace

External resources

*Building Hardened Docker Images from Scratch with Kubler