Running elastic search with 2, 3 and 4 nodes. Configuration is generic recommended to insure the cluster can respond to requests if all but one node goes down.
For 3 and 4 node configurations, I'm seeing some serious issues. There are multiple nodes electing themselves as master. What's worse, sometimes those master nodes only know about N-1 of the other nodes.
My cluster is named the same in every configuration file, and each node was given a unique name as well. The query performance is taking a hit because multiple nodes think it's only an N-1 cluster, and it's impossible to guess which node is going to decide to elect itself as a dumber master.
Any ideas?
Aucun commentaire:
Enregistrer un commentaire