Performance is low when using cluster mode

Elasticsearch Version

7.4.0

Java Version

from official docker image

OS Version

from official docker image

Problem Description

I have set up an Elasticsearch cluster using the official Elasticsearch image. I have configured three nodes, each with 2 CPU cores and 8 GB of RAM. I have also isolated the CPU resources for each container using Docker to avoid CPU resource conflicts.

I conducted a test using a dataset in the order of 100,000 records. Initially, I set the shards count to 1 and the replica count to 0, and the throughput was measured at 1000 requests per second (r/s). However, when I increased the replica count to 2, the throughput significantly decreased to 300 r/s. I also tried to set the shard count to 3 and the replica count to 1, but the throughput still experienced a significant drop.

I attempted to increase the number of CPU cores per node to 4, but performance did not improve. How can I sustainably enhance the performance of my Elasticsearch cluster, such as by increasing the number of nodes and configuring the appropriate shard and replica settings, for easier maintenance in the future?

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