I have 3 nodes and indices with 3 shards and one replica. I see some of my old indices have primary shards on the same node. I believe this happened during midnights, when indice was created, but some node was down. Now I can manually move shards to different nodes using cluster reroute, but I thought this should be done automatically. But this is not happening, although cluster.routing.allocation.enable is set to all. So how to achieve this? Should I use curator? And am I right thinking that having all primary shards on same node is slower than having them spread across cluster?
Thank you very much for help.