We have setup a 13 node elasticsearch cluster (3 master, 10 data nodes) such that 7 of the data node have an attribute A=1 and 3 nodes have an attribute A=2
This is because we want to ingest a lot of data into the cluster without impacting existing indices on it
Now that we have the already existing indices separated into the 3 nodes marked A=2, we were planning to ingest as fast as we can onto the 7 nodes marked A=1
We won't be ingesting via client nodes, we'll directly access the 7 data nodes
Do you think this is a good approach?
Also, once we have migrated all the data, we'll be removing the indices setting such that everything gets spread across all the 10 data nodes. This means, there'll be rebalancing. In shard allocation filtering, does the cluster ever go to yellow? How do we throttle the rebalancing so as to not affect the entire cluster? There were a few setting like cluster.routing.allocation.cluster_concurrent_rebalance
but the doc says this does not take into consideration rebalancing due to shard allocation filtering. What are my options