I have hot warm architecture setup in AWS, which is based on instance type (i2 - hot, d2 - warm), and I want to perform a recovery from an old snapshot. I applied index creation template (a common one for all indexes) to have index routing allocation require as d2 type (warm) instance and started restore with include_global_state as false so that old index templates (stored with the snapshot) are not given preference. However, shards got allocated evenly across hot and warm nodes.
After that, I ran curator job to force merge and then allocate the indices to warm nodes, which initiated the process and started moving the shards to the warm nodes.
How does the currently applied index templates work in case of snapshot recovery with include_global_state set as false?