We have a small ElasticSearch 2.4 cluster containing 2 Data Nodes and multiple Indicies.
The indicies have 2 primary shards and 1 replica each.
If we create a new index using our default template, it creates one primary shard on node1 and the other on node2.
The replica shards are assigned the opposite way. So node1 contains primary0 and replica1 and node2 contains primary1 and replica0.
Yesterday i had to restart node1, now node1 contains only replica0 and replica1 and node2 contains only primary0 and primary1. The CPU usage on node2 suffers from this, but i dont really know why because replica shards should serve read requests too.
Here you can see what we have (first 2 indexes), and what we want (last index)
How can i reassign a replica shard to be primary?