Coordination Node Client request stats

Hi, In our Elasticsearch cluster, there are three coordinatin nodes (dedicated) which are forwarding client search requests to Hot Nodes. There is no load balancer before Coord Nodes. They are receiving client request directly from application. I want to find out which Coord Node is receving how much Client search requests, so that the client search request load distribution among them can be monitored.
In node/stats, I am not able to find out the parameter, which is logging this.

The Monitoring functionality will show you this.

Tx, Which parameter in Kibana's Node monitoring. If it is System Load, which is showing some variation , but it'll will contain indexing request load also alongwith client search request. If, Request rate , but it shows the amount of search operation which is not the case.
It will be helpful if some specific monitoring matrix can be suggested. Or any other method ?

Also , How to achieve the equal load distribution among all coord Nodes. Do I need to use any seperate load balancer before Coord Nodes.

Kindly respond

Why do you need to achieve even load across the nodes? In a highly available setup a subset of the nodes need to be able to handle the full load in case one of the nodes were to go down.

Having a load balancer in front of the nodes may help as I am not sure if there is a pattern to how the clients establish connections. Also be aware that most clients establish long-running persistent connections for efficiency, so may not rebalance very quickly. Limiting the lifetime of a connection may help address this but you do not want to establish a connection per request due to the overhead. How to do this will depend on the client(s) used.

Sometimes the uneven load on EC nodes may overwhelm a particular EC node resulting in heap utilization reaching 100% and making that node non-responsive. Under these conditions, the queries to this node and response will be lost. How the client API sending the queries to all EC nodes will know this status and how the redundancy or failover will be achived.

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