This is a small collection of quickly put-together ansible scripts to set up Gitlab runners with the shell executor environment properly configured to build my KVM Conjurer project.
Each script can be invoked with the ansible-playbook command. The configuration of the Python 3 build environment on the Debian VMs I use requires specifying the Python 3 interpreter:
$ ansible-playbook -i inventories/some-inventory.yml \
-e 'ansible_python_interpreter=/usr/bin/python3' \
configure-conjurer-buildenv.yml \
To safely store the tokens used in registering a Gitlab runner, I use an ansible-vault encrypted file:
$ ansible-vault encrypt --vault-id ID@vaultpasswdfile secrets/keys.yml
$ ansible-playbook -i inventories/some-inventory.yml \
-e secrets/keys.yml \
--vault-id ID@vaultpasswdfile \
configure-gitlab-runner.yml \
where secrets/keys.yml is of the format
gitlab_url=http://your.url.com
gitlab_api_token=YOUR_API_TOKEN
gitlab_runner_token=YOUR_RUNNER_REGISTRATION_TOKEN
As mentioned above, the script configure-gitlab-runner.yml uses the default python on Debian 10 (Python 2.7). The python-gitlab library version availabe is 1.15. This version is compatible with the ansible.community.gitlab_runner module.
The version of python-gitlab availabe on PyPI for Python 3.7 is 2.5 or so. This is not compatible with the Ansible module. I'm investigating whether a patch would be easy to produce or if the problem goes deeper than one broken instance of subscript notation.