I have a 3 node cluster with 5 primary shards and no replicas. My environment feeds bulk insert requests into the ES cluster. However, when I look at _cat/thread_pool, I see that one of my nodes always has its bulk threads active with quite the buildup in the queue. The other two nodes sometimes have threads active and no buildup in the queue.
My shards are equally distributed across the cluster.
How can I get my cluster to evenly spread out the bulk requests?