// the find
prometheus-community/elasticsearch_exporter
Elasticsearch stats exporter for Prometheus
A Prometheus exporter for Elasticsearch and OpenSearch cluster metrics — node stats, index stats, shard state, snapshots, ILM, and more. It sits in front of your ES cluster and scrapes the various `/_stats` and `/_nodes` endpoints on each Prometheus scrape. If you're running ES and already have Prometheus, this is the obvious bridge.
The multi-target `/probe` endpoint added in v2 is genuinely useful — one exporter instance can cover a fleet of clusters with different auth methods (basic, API key, AWS SigV4, mTLS), configured from a YAML file. The test suite uses fixture JSON from real ES versions going back to 1.7.6, which catches response-shape regressions across major versions without needing a live cluster. The RBAC privilege table in the README is exactly what you need when your security team asks why Prometheus needs cluster:monitor. The included Grafana dashboard and alerting rules mean you're not starting from scratch.
The default 5s ES timeout and the warning about scrape load in the README are telling — on a large cluster with `--es.all --es.indices`, each scrape hits the ES master with a full cluster stats query, and there's no internal caching between scrapes. If Prometheus scrapes at 15s intervals and each scrape takes 10s, you have a bad time. Multi-target mode is still marked beta and the auth module config format isn't stable yet. There's no support for ES 8.x cross-cluster search metrics or the newer `_cluster/stats/nodes/` decomposed format, so on very large clusters you're still pulling one giant JSON blob. Shard-level metrics (`--es.shards`) on a 500-shard cluster will produce so many time series that you'll hit Prometheus cardinality limits before you notice.