I have 2 availability zones in AWS.
In each AZ, I have 2 auto scaling groups –
- 1 node of i2 (with instance storage) machine and 2) 2 Nodes d2 (EBS volume) machines
I use instance type in the elasticsearch.yml as the tag for hot warm.
I have a curator job that force merges the indices and then performs an allocation require action.
By default I have 2 shards per indices (<100 GB)
As expected, new allocation of shards happen to the hot nodes (i2) as defined in the index template.
The issue is, when the curator job is executed, only the replica shards move to the warm nodes (d2) !!!
What is causing this?