Following the advice for deploying a cluster using bootstrap: Deployment considerations
We have new-cluster-enabled set to off in our cluster in a Kubernetes deployment using discovery-api. We also have configured the split brain resolver with “keep majority” strategy.
If the split brain resolver decides to down all nodes, then the cluster is not able to form again due to new-cluster-enabled being set to off. This requires monitoring to alert operators the need to manually recover from this incident.
Is there a way to automatically recover from this scenario or is it not recommended to allow downing all nodes? These two recommendations from the docs seem to conflict with each other in this failure scenario