Approx sizing for indexes and shard counts per data node

I know what I'm asking can be subjective but I'm wondering if I have say 5 data nodes with access to 29gb Ram in the JVM (64gb of physical ram), with 5 data nodes, what is the best way to estimate when I need to add a new data node. Like is there a. best practice shard count per data node of this size i should say within or document count that this cluster happily manage. Trying to do some growth forecasting of when a new data node should be added. Is there like a forumula that is used to forecast these things and such. These index documents are not extensive in size, (like 30 fields or so). Just trying to come up with some rough guessing of when to have the clients to look into horizontally expanding portions of the cluster PRIOR to the cluster going red.


Usually when you reach 6-700 shards per node, or if you are consistently >75% heap use.

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.