Just curious how do you route shards from hot to warm nodes and prevent same shard allocation on a physical server in a Vertical scaling model where each physical servers
- 2 or more hot ES nodes with data on SSD
- 2 or more warm ES nodes with data on HDD
cluster.routing.allocation.same_shard.host which worked for some indexes but failed for others because shards on hot ES nodes refused to move to warm ES nodes on the same physical server because they think they are on the same physical server (which is actually true). However, what I want to prevent is that same shards (primary and replica) should not allocate on hot nodes or on warm nodes on the same physical server, yet allow shards to move from hot to warm as if they are on a separate physical server.
I believe https://www.elastic.co/guide/en/elasticsearch/reference/2.3/allocation-awareness.html is the solution for such deployment, but would appreciate more inputs from the community.