Help figureing out the thread pool distribution

(Stecino) #1


I have 15 node cluster, 3 masters, 2 queries and 10 data nodes. At the moment, I am running into some ingestion issues by this application, while other indices are ok. I have been playing with the index.queueSize and it has gone up to 4500, at which point I have no index rejections but queue remains pretty high, and only spread among two data nodes. Is there a way to tell ES to spread index.queue distribution among all 10 data nodes? Hopefully, I was able to explain it right. At this screen capture it's distributed to 4 nodes looks like

(Mark Walkom) #2

Which nodes have the high count, the clients?

(Stecino) #3

Data nodes

(Mark Walkom) #4

Are you sending indexing requests to the client nodes or to the data nodes?

(Stecino) #5

In this particular case, it's java clients that are sending requests to data nodes.

(Mark Walkom) #6

Then you probably need to check that the clients have load balancing setup correctly.

(Stecino) #7

Is there a best practices guide in load balancing the data nodes from the client? Also, they call some groovy stuff also. I know that groovy isn't recommended by Elastic. Is there anything else other than groovy that can be recommended?

(system) #8

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.