Hi, I have an elastic 2.3 legacy cluster that I have to scale up to prepare for a significant seasonal load.
I observe that there are 2 indices on the cluster with the following configs:
Index P:
"number_of_shards": "4"
"number_of_replicas": "4"
green open P 4 4 591635702 3550680 357.8gb 71.6gb
Index S:
"total_shards_per_node": "1"
"number_of_shards": "6",
"number_of_replicas": "4",
green open S 6 4 252376492 51806564 788.2gb 160.7gb
Current size of active nodes that actually keep the data is 31. Since I have to scale this up my idea is to add nodes until every shard of every index had its own node. Some sources say that if I have several shards of the same index on the same nodes, the querying happens in a serial fashion which may slow things down. I don't know if the same is true for different indexes, but I would assume so (please let me know).
Current node type : 16 vCPUs, 60 GB memory, 375 GB SSD
What other action I could potentially take in order to improve performance? Giving the cluster an additional 19 nodes may be pretty expensive, I want to make sure that I explore other options before I do that.
Let me know if there are other settings that might be relevant here.
Thanks!