new-cluster-enabled=false doesn't work well with SBR down-all. It doesn't have to be the down-all SBR strategy but there are also other situations when SBR will resort to downing of all nodes and a new cluster must be formed.
SBR and Cluster Bootstrap are two independent processes so it's not easy to say that new-cluster-enabled should be false unless SBR downed all nodes before the bootstrap.
In Kubernetes it should be safe to use required-contact-point-nr > 1 instead of new-cluster-enabled=false. I haven't even seen any problems with required-contact-point-nr=1.
We should at least document this caveat.
new-cluster-enabled=falsedoesn't work well with SBR down-all. It doesn't have to be the down-all SBR strategy but there are also other situations when SBR will resort to downing of all nodes and a new cluster must be formed.SBR and Cluster Bootstrap are two independent processes so it's not easy to say that
new-cluster-enabledshould be false unless SBR downed all nodes before the bootstrap.In Kubernetes it should be safe to use required-contact-point-nr > 1 instead of
new-cluster-enabled=false. I haven't even seen any problems with required-contact-point-nr=1.We should at least document this caveat.