I have been unable to figure out (or find in documentation) what exactly happens when elasticsearch data node starts rejecting search requests, so I am hoping you guys could help me to understand.
My main concern is what is the expected response (ideally HTTP status or response JSON body) to search request that's rejected?
You see, some of our data-nodes started logging large quantities of rejected requests under stress test, but the REST client sending search requests did not log any HTTP states other that 200.
That leaves me thinking that the search rejections are on data node level only and rejected search request are retried by coordinating node either to the same node later or to other node holding overwhelmd shard's replicas?
How many times is coordinating node retrying these requests to overwhelmd shard?
And what is the response to the original HTTP search request on the HTTP level?
I hope I described my questions sufficiently and that someone will be able to answer them. If not, I will gladly provide more details.
With kind regards,