Cluster becoming red during rolling upgrade


(Bittu Sarkar) #1

As per the rolling upgrade process documented here, it is advised to set the value of "cluster.routing.allocation.enable" to "none" before bringing a node down for servicing and set it back to "all" once the node is serviced. The problem is if a new index gets created while the node is being serviced, the cluster goes to RED state. This is a real scenario when xpack creates new monitoring indices every day at 00:00 UTC. Of course it will go back to YELLOW and then to GREEN state once allocation is enabled, but we would never want the cluster to be in RED state. Instead of setting it to "none", doesn't it make sense to set it to "new_primaries"? That way it is guaranteed that the cluster will not go into RED state when there is no problem with the system. Please advise if there is anything wrong with this approach. If not, should we change the official documentation as well?


(system) #2

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.