Due to storage limitations, we are unable to backup the entire mlflow artifact store, as a minimum, we will backup the mlflow_db which should allow us to retrain and reproduce any artifacts from a particular run.
In addition to this, we should backup the model registry to ensure deployed models can be restored quickly.