gitlab-runner in vagrant

Example scripts to run different GitLab runners as virtual machines under vagrant, as disposable, on-demand workers.

Shell scripts provided so that you really do not need anything else. This can be easily converted into kickstart/preseed/cloudinit scripts.

Notice

This is a quick proof of concept and is far from any production use. There are already puppet/ansible modules/roles existing in the world.

Providers

• virtualbox (not tested, but should work)
• libvirt
• LXC (Centos only), but it is deprecated, use vagrant-lxd
• LXD 3.x with vagrant-lxd

Flavours

• Centos 7.4 + gitlab-runner as docker executor (basic)

• Centos 7.4 + gitlab-runner as shell executor, with docker, java and gcc++

• Ubuntu 16.04 + gitlab-runner as docker executor (basic)

• Ubuntu 16.04 + gitlab-runner as shell executor, with docker, java and gcc++

Centos installs new kernel mainline from elrepo. Ubuntu installs kernel HWE LTS.

OpenJDK 8, headless, but no ant/maven.

Preparing

• ensure you have vagrant installed, verion 2.2.5
• ensure you have vagrant plugin installed for given provider and it is properly configured, pro tip: read docs for given provider really carefully!
• ensure you have vagrant plugin vagrant-reload

vagrant plugin install vagrant-reload

• (optional) ensure you have proper plugin installed for given providers, for example vagrant-libvirt, vagrant-lxc, vagrant-lxd

• ensure you have configured system to be able to use given provider, refer to documentation details provided by given plugin (especially for LXD)

• go to your GitLab install, find project

• go to Settings (on the bottom left), CI/CD, Runners settings, click expand

• there is a section 'Specific Runners', with URL and token in red, copy it.

Adjusting setup

• go into specific directory

• copy .secrets.dist to .secrets

• edit .secrets and update runner name, URL, project token to register and tags

• execute command in shell, to load env vars:

. .secrets

• after command above executed, start vagrant:

vagrant up --provider=<providername>

• in the end of vagrant execution you should see something in red color (yeah...):

vm-name: Registering runner... succeeded   runner=SOMETHING
vm-name: Runner registered successfully. Feel free to start it,
         but if it's running already the config should be automatically reloaded!

• go to the GitLab project, and check if in Runner settings you see registered runner.

• check your GitLab project and edit .gitlab-ci.yml to use proper tags in the project depending on the build/step and so on.

Rebuilding machine

Pretty obvious

vagrant destroy -f
vagrant up --provider=<providername>

Remember to go to GitLab CI and remove dead workers.