prometheus-metrics plugin for ElasticSearch

Elasticsearch exporter plugin returns internal status data in a Prometheus metrics format for monitoring and alerting purposes. Comes together with rules for alerting It can deliver basic information about cluster, indices and JVM status in an asynchronous way. Just add it as a target endpoint and start collecting data from the internal status metrics of elasticsearch database.

Installation

To install this plugin just add it into your version of ES. Example for 0.15.0 version for ES 7.4.2:

bin/elasticsearch-plugin install https://github.com/jsuchenia/elasticsearch-prometheus-metrics/releases/download/0.13.3/prometheus-metrics-0.15.0-7.4.2.zip

and register it in your prometheus & add rules!!. Example docker image can be run using a command:

  docker run -p 9090:9090 jsuchenia/prometheus-elasticsearch:v2.15.1-0.15.0

and visit http://localhost:9200

Features

Run of variety of Elasticsearch versions without any dependency - just pure asynchroneus java - so it's not blocking threads with sync calls.

After installation it will expose few HTTP endpoints:

• /_prometheus/node - with details about node including:
  • JVM - most of metric aligned with client_java code
  • OS - metrics about an operating system
• /_prometheus/cluster - Cluster and indices status
• /_prometheus/cluster_settings - Cluster wide settings to check global run-time values as metrics
• /_prometheus - Overall status - includes all metrics from above endpoints

To use it just add target URL to your prometheus: http://elasticsearch.domain.com:9200/_prometheus

Console

Together with plugin & rules you can find here a console templates. When extracted into /etc/prometheus/ (two directories: console and consoles_lib) you will find out additional button with few simple dashbards about cluster and nodes

Rules

Simple rule to monitor cluster health (v1 version):

ALERT ElasticSearchClusterStatus
  IF es_status > 0
  FOR 1m
  LABELS {
    severity="critical"
  }
  ANNOTATIONS {
    description="{{$labels.instance}} reports non-healthy status of ElasticSearch cluster",
    link="http://prometheus.host:9090/alerts",
    summary="NonHealthy cluster on {{$labels.instance}}"
  }

Few rules that we use are located in rules file:

• Prometheus 1.X version: elastic-rules.rule
• Prometheus 2.X version: elastic-rules.rule.yml

Rules assumptions

• Cluster can be usntable for 10 minutes - after that we will send severity=warning alert
• Cluster can be unavailable for at most 1m - after that we will send severity=critical alert

Hints for rules transformation:

You can convert rules by executing a command:

docker run --rm -it -v "`pwd`:/prometheus" --entrypoint /bin/promtool prom/prometheus update rules elastic-rules.rule

And validate them via command:

docker run --rm -it -v "`pwd`:/prometheus" --entrypoint /bin/promtool prom/prometheus check rules elastic-rules.rule.yml

Supported versions by latest releases (only 7.3.X and above)

• 7.5.1
• 7.4.2
• 7.4.1
• 7.4.0
• 7.3.2

Supported prometheus Releases

• 2.11.1

Supported versions in previous releases

• 7.3.0
• 7.2.1
• 7.2.0
• 7.1.1
• 7.1.0
• 7.0.1
• 7.0.0
• 6.8.1
• 6.8.0
• 6.7.2
• 6.7.1
• 6.7.0
• 6.6.2
• 6.6.1
• 6.6.0
• 6.5.4
• 6.5.3
• 6.5.2
• 6.5.1
• 6.5.0
• 6.4.2
• 6.4.1
• 6.4.0
• 6.3.2
• 6.3.1
• 6.3.0
• 6.2.4
• 6.2.3
• 6.2.2
• 6.2.1
• 6.2.0
• 6.1.4
• 6.1.3
• 6.1.2
• 6.1.1
• 6.1.0
• 6.0.1
• 6.0.0

Generated .zip plugins are available at Releases section of this project (in a future in a maven repository)

Security

Currently this plugin do not provide any REST access rules, consider to use other plugins like Search Guard