I have 6 nodes and 6 primary shards for each index plus replica.
I want to be sure that each index primary shard got own node.
How I understand, this can be controlled by allocation.balance
"cluster.routing.allocation.balance.shard": 0.05, "cluster.routing.allocation.balance.index": 0.95, "cluster.routing.allocation.balance.threshold": 0.95,
But looks like this settings work with primary and replica shards. As result I am getting the following:
Here one problem. Due ILM policy after few days replica will be dropped. And after replica dropped - rebalance will start and move shard 0 or 5 to node es-logs-1-1. Meaningless operation. Is any way to do not allow this kind of spread?