Hello, I am now planning an Elasticsearch 5.4 setup which will be used for data migration. The data consists of unstructured text and meta data.
The projected size of the index for all the migrated data is 750GB.
Based on 40GB per shard, around 20 primary shards are needed. I plan for 1 set of replicas so there will be a total of 40 shards.
For each VM, there will be 1 Elasticsearch node.
For best query performance, should I focus on getting more VMs with less shards per node or fewer VMs but with more shards per node?
This Elasticsearch setup will only be indexed once. Once all migrated data has been indexed, the setup will only be used for searches.
Also, if I follow the formula of n/2+1 where n is the number of master eligible nodes, does it matter whether I have an even or odd number of master eligible nodes? (I plan to have 4 VMs, hence 4 master eligible nodes).
Thank You