Elasticsearch Cluster Monitoring Metrics Supported by the Kernel

P99 latency for search requests

Open-source Elasticsearch provides only the average latency metric when monitoring responses to search requests. This may not accurately reflect the actual search performance of a cluster. To improve on this, the P99 latency metric is added in CSS to monitor the 99th percentile latency of each cluster.

Elasticsearch 7.6.2, Elasticsearch 7.10.2

Monitoring P99 Latency

HTTP status codes for the cluster

When you access Elasticsearch through HTTP, you receive HTTP status codes in response to your requests. The native open-source Elasticsearch does not collect statistics on these status codes. To improve on this, HTTP status code monitoring is added in CSS, allowing you to monitor HTTP status codes and get a sense of how the service is running.

Elasticsearch 7.6.2, Elasticsearch 7.10.2

Monitoring HTTP Status Codes

Cluster indexes

Index monitoring helps to monitor the index usage and trends of a cluster, allowing users to handle potential risks in a timely manner to ensure cluster reliability. Index usage statistics are collected and stored in the cluster's monitoring index (index name: monitoring-eye-css-[yyyy-mm-dd]. By default, these monitoring indexes are retained for one week before being permanently deleted upon expiration.

Elasticsearch 7.6.2, Elasticsearch 7.10.2

Configuring Index Monitoring for an Elasticsearch Cluster