For an index with replica shard(s), does the docs.count field given by /_cat/indices include docs on the replica shard or only docs on primary shards?
Say I want to get even splits of an index by using the scroll API with a given size. If in this case docs.count would include docs on the replica shards, would scrolling over portions of the total docs returned by _cat/indices yield copies of data across different scrolls as it would scroll over the primary document on the primary shard and then again over the replica shard?
Example for clarification: Say we have index1 with 1 primary shard, 1 replica shard, and _cat/indices returns 1000 for docs.count. Now I want to divide index1 into 4 sections, so I use the scroll API with a size of 250.