We maintain a hot-warm architecture and I have noticed when I'm trying to move the shards from hot nodes to warm nodes, during the initial phase the a few shards would be moving from hot to warm, but this used to trigger a rebalance within hot nodes and the shards which at the end has to be moved from hot to warm are also moving within hot nodes. Considering the network bandwidth this slows down our relocation time a lot.
Is it possible to set node specific settings like I want to use 0 for
node_concurrent_incoming_recoveries on hot nodes and 20 on warm nodes
I've tried to find documentation regarding this but couldn't find any thing that could help in this case.