This page mainly shows you how to build and use your own custom Execution Environment.
When the Job Template is launched on AWX, the playbook runs on the Execution Environment (EE), which is a containerized environment completely isolated from your K3s host.
The default EE, quay.io/ansible/awx-ee:latest, has some typical collections, Pip modules, and RPM packages by default, but sometimes your playbooks require additional (non-default) collections, modules or packages.
In such cases, you need to add Ansible collections, Python packages, and RPM packages to the EE. There are two ways you can choose to achieve this.
- Build and use your own Execution Environment
- This method is used when additional Pip or RPM packages are required in addition to the Ansible collection.
- Follow the whole of this page to build and use a custom EE.
- Place
collections/requirements.yml
in your project- This is useful if you simply need additional Ansible collections only, and no additional Pip or RPM packages are required. However, this method cannot be used if your project type on AWX is Manual.
- See Use Ansible collections without custom EE at the bottom of this page for instructions.
To build a custom EE, there is a tool called Ansible Builder. You can build your own custom EE with any Ansible collections, Python packages, and RPM packages added.
This repository includes ready-to-use files as an example to use Ansible Builder. You can clone my repository to start with my ready-to-use example files.
cd ~
git clone https://github.com/kurokobo/awx-on-k3s.git
cd awx-on-k3s/builder
- CentOS Stream 9 (Minimal)
- Python 3.11
- Docker 25.0.4
- Ansible Builder 3.1.0
You can install Ansible Builder by using Pip.
python3 -m pip install ansible-builder
At least, the file execution-environment.yml
is required to build an EE.
This repository contains execution-environment.yml
as a minimal ready-to-use example. This file is made to achieve the following requirements.
- Use
quay.io/centos/centos:stream9-minimal
as the base image - Use Python
3.11
as the Python interpreter - Use Ansible
2.15.*
and Ansible Runner2.3.*
to run playbooks on the EE - Add RPM Packages that are listed in
dependencies/bindep.txt
for basic remote connection and debugging - Add Python packages that are listed in
dependencies/requirements.txt
- Add Ansible collections that are listed in
dependencies/requirements.yml
- Customize the configuration for
ansible-galaxy
(additional_build_files
andadditional_build_steps
)- Add a customized
ansible.cfg
to the context (additional_build_files
) - Place the customized file as
~/.ansible.cfg
for the stage thatansible-galaxy
is invoked (prepend_galaxy
underadditional_build_steps
)
- Add a customized
- Run additional commands during build steps (
additional_build_steps
)- To allow the hard-coded interpreter name (
python3
) to be passed by AWX, thealternatives
command is appended underappend_base
to make the binary/usr/bin/python3.11
executable as commandpython3
- To allow the hard-coded interpreter name (
Note that since this example uses the *-minimal
image as the base image and added only a few packages for SSH connection and debugging, there should still be missing packages and modules for some modules and collections.
You can review and modify execution-environment.yml
and any files referenced from this file to suit your requirements.
The syntax of requirements.yml
is the same as for Ansible Galaxy. The syntax of requirements.txt
is the same as for Pip, and bindep.txt
is for Bindep.
Other customizations are possible besides this. Refer to the official Ansible Builder documentation for details.
Once your files are ready, run ansible-builder build
command to build an EE as a container image according to the definition in execution-environment.yml
. Specify a tag (--tag
) to suit your requirements.
ansible-builder build --tag registry.example.com/ansible/ee:2.15-custom --container-runtime docker --verbosity 3
Below is an example output of this command.
$ ansible-builder build --tag registry.example.com/ansible/ee:2.15-custom --container-runtime docker --verbosity 3
Ansible Builder is generating your execution environment build context.
File context/_build/requirements.yml will be created.
File context/_build/requirements.txt will be created.
File context/_build/bindep.txt will be created.
Creating context/_build/configs
File context/_build/configs/ansible.cfg will be created.
File context/_build/scripts/assemble will be created.
File context/_build/scripts/install-from-bindep will be created.
File context/_build/scripts/introspect.py will be created.
File context/_build/scripts/check_galaxy will be created.
File context/_build/scripts/check_ansible will be created.
File context/_build/scripts/pip_install will be created.
File context/_build/scripts/entrypoint will be created.
Ansible Builder is building your execution environment image. Tags: registry.example.com/ansible/ee:2.15-custom
Running command:
docker build -f context/Dockerfile -t registry.example.com/ansible/ee:2.15-custom context
#0 building with "default" instance using docker driver
#1 [internal] load build definition from Dockerfile
#1 transferring dockerfile: 3.14kB done
#1 DONE 0.1s
...
#5 [base 1/7] FROM quay.io/centos/centos:stream9-minimal@sha256:08da5b444b6f7bdc1e4253e36fd3bd52d712d9c19c1b9e426fd733c5865c9d30
...
#12 [galaxy 1/7] ADD _build/configs/ansible.cfg ~/.ansible.cfg
...
#13 [builder 1/7] WORKDIR /build
...
#18 [final 1/9] RUN /output/scripts/check_ansible /usr/bin/python3.11
...
#35 exporting to image
#35 exporting layers
#35 exporting layers 1.1s done
#35 writing image sha256:34147e320cd92bf40c545c785e7718dc65936d1b68179f0620b329a286095064 done
#35 naming to registry.example.com/ansible/ee:2.15-custom done
#35 DONE 1.1s
Complete! The build context can be found at: /home/********/awx-on-k3s/builder/context
Once the command is complete, your custom EE image is built and stored on Docker or Podman.
$ docker image ls
REPOSITORY TAG IMAGE ID CREATED SIZE
registry.example.com/ansible/ee 2.15-custom 34147e320cd9 20 seconds ago 285MB
You can use your EE in your AWX or Ansible Runner.
To use your EE in AWX, in typical use cases, your EE should be stored on some container registry and should be pulled by AWX, but technically you can use your EE with your AWX without any container registry.
Simply you can push your EE image to some container registry. Any registry can be acceptable. If you want to deploy your own private container registry, refer additional guide on this repository.
$ docker push registry.example.com/ansible/ee:2.15-custom
The push refers to repository [registry.example.com/ansible/ee]
...
2.15-custom: digest: sha256:b0d4638a788a74d893075f5600a00e5aef51ada67e46bde2071615fc936ef0d8 size: 3661
Then you can specify registry.example.com/ansible/ee:2.15-custom
as your own custom EE in AWX. Specify registry credentials if your container registry requires authentication.
The EEs in AWX are just Pods on Kubernetes. When executing a Job Template in AWX, AWX instructs Kubernetes to create a Pod using the specified EE image. Of course Kubernetes requires the specified container image when creating the Pod. So while creating the Pod, Kubernetes usually pulls the specified image from the container registry.
But this happens only if the image is not cached on the container runtime for Kubernetes. If Kubernetes has the cache of the specified image, Kubernetes never pulls the image from the container registry (Technically, this is the default behavior and stated as imagePullPolicy
is IfNotPresent
).
This means that if your Kubernetes has all the EE images you need in its cache in advance, you can use any EE from AWX without any container registry. To achieve this, you should export your EE image from Docker or Podman, and import it to containerd that the container runtime for K3s.
# Save your EE image as Tar file
docker save registry.example.com/ansible/ee:2.15-custom -o custom-ee.tar
# Import the Tar file to containerd
sudo $(which k3s) ctr images import --compress-blobs --base-name registry.example.com/ansible/ee:2.15-custom custom-ee.tar
Ensure your imported image is listed.
$ sudo $(which k3s) crictl images
IMAGE TAG IMAGE ID SIZE
...
registry.example.com/ansible/ee 2.15-custom 34147e320cd92 295MB
...
Now you can specify registry.example.com/ansible/ee:2.15-custom
as your own custom EE in AWX without any container registry and any credentials.
In AWX, you can change the policy of pulling the image in Edit
page of your EE. The default Only pull the image if not present before running
is ok, but to be safe you should specify Never pull container before running
.
Refer to the guide for Ansible Runner on this repository.
Some additional information around Ansible Builder and EE.
The Dockerfile
is generated and stored under the context
directory once your ansible-builder build
command has completed.
$ cat context/Dockerfile
ARG EE_BASE_IMAGE="quay.io/centos/centos:stream9-minimal"
...
# Base build stage
FROM $EE_BASE_IMAGE as base
...
# Galaxy build stage
FROM base as galaxy
...
# Builder build stage
FROM base as builder
...
# Final build stage
FROM base as final
...
If you want to create Dockerfile
without building an image, you can use the ansible-builder create
command.
$ ansible-builder create --verbosity 3
Ansible Builder is generating your execution environment build context.
File context/_build/requirements.yml will be created.
File context/_build/requirements.txt will be created.
File context/_build/bindep.txt will be created.
Creating context/_build/configs
File context/_build/configs/ansible.cfg will be created.
File context/_build/scripts/assemble will be created.
File context/_build/scripts/install-from-bindep will be created.
File context/_build/scripts/introspect.py will be created.
File context/_build/scripts/check_galaxy will be created.
File context/_build/scripts/check_ansible will be created.
File context/_build/scripts/pip_install will be created.
File context/_build/scripts/entrypoint will be created.
Complete! The build context can be found at: /home/********/awx-on-k3s/builder/context
If you simply need additional Ansible collections only, and no additional Pip or RPM packages are required, this method is the simplest way. However, this method can't be applied for Manual type projects. If you want to add your project as Manual type, you should build custom EE.
The procedure is quite simple, create and place collections/requirements.yml
in your project root including the list of the collections that you want to use.
The format of collections/requirements.yml
is the same as the requirements.yml
for ansible-galaxy.
If collections/requirements.yml
is present in your project, AWX will install the collections accordingly while updating project.