It looks like you are overloading your cluster, as the bulk threadpool is getting full and rejecting requests.

Do you have Monitoring enable?

---