If it is just the current master node that is busy it could be large, complex or frequent updates to the cluster state may be the cause. Run the hot threads API to find out. This could be caused by shard creation or relocation and/or mappings that are frequently dynamically updated.