Load balancer for elasticsearch cluster

  • Elasticsearch version - 5.1.2
  • 16GB of max heap configured
  • Installed and used enterprise modules, if any - No
  • JVM version and operating system version - openjdk version "1.8.0_292" , CentOS
  • Other installed Elasticsearch or Kibana plugins, if any - None
    Configuration: Google Loadbalancer -> 3 Instances -> 3 data and master nodes (master and data on same instances)

Hi, my setup right now has 3 instances with both data and master on each of them.
I use Google internal load balancer "before" instance group of this instances.
Recently I have configured a NewRelic monitoring for this cluster and I have seen this picture

The interesting part is Open http connections and READ IO Operations that it shows on es-node-1.

I wonder what would be a best practice or a suggestion having this configuration and what is the best way for connecting an application to cluster? Directly to one of nodes? Using load balancer same way I use it now ? DNS with round robin ? Any other 3rd party LB solution for ES?

I guess it depends on how the load balancer is setup to talk to the cluster?

5.X is very much EOL, please upgrade as a matter of urgency.

Hi Mark, thanks for reply,
The load balancer is an Internal TCP load balancer Pass-through.
I'm trying to figure out what would be the best way to connect an application to elasticcluster?

That depends on the application.
If you are using an official client then it will handle the load balancing etc. Otherwise, HTTP level would make sense.

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