I would suggest combining Pod Disruption Budgets PDBs for your RabbitMQ pods which will ensure at least your two RabbitMQ pods arealways running for example and the managed nodes rolling update strategy: