I have implemented a hot-warm architecture in my cluster. I use Curator to re-tag the desired indices from 'hot' to 'warm' at a set time but I have noticed recently that the migration of the tagged indices is taking longer and longer (as the index sizes are growing). Which cluster setting, if any, controls how much bandwidth is available for moving an index from a hot node to a warm one? Everything I've seen in the docs relates to bandwidth during recovery of indices, not just index migration and re-balancing.
Thanks in advance.