After we migrate production change-prop to k8s, we would be left with a working deployment in beta cluster. For the rest of the services, it seems like beta deploys were replaced with a simple docker container runner. This would be enough for change-prop, but the problem would be that change-prop config is extremely complicated, so duplicating the config for beta cluster would not be a satisfying experience.
Instead, we should experiment with reusing the k8s config in beta. Perhaps, it could involve some manual steps, but the important part is to have the source of truth for the config shared. How would it work:
- We would have special 'beta' values.yaml
- We would do helm template to render the config. It will render much more then just the config itself for us, be it would be easy enough to trim the output to get the config.
- We would deploy the rendered config on beta note automatically.
This could be done via some updater script calling to k8s api from the beta node.