Problem Description

Bulk Index request which change the mapping of the index can cause put mapping requests. These requests need to be send across to the master for processing and all this while other bulk requests are held up on the co-ordinator node for processing. This is a problem if there are high priority items with priority URGENT already queued up on master(too many shard-started events due to fluctuating nodes)as this would cause a surge in memory pressure for the coordinating nodes and bulk indexing to get slowed down