Monitoring

The OpenSearch charm integrates with the Canonical Observability Stack (COS) to provide infrastructure and cluster health monitoring through metrics, dashboards, alert rules, and logs. COS uses Prometheus for metrics collection, Grafana for visualization, Loki for log aggregation, and Alertmanager for alerting. This enables monitoring of OpenSearch cluster performance, resource utilization (CPU, memory, disk I/O), cluster statistics (node health, shard allocation, indexing rates), and operational health.

The way telemetry reaches COS depends on the charm variant:

• On VMs, the opensearch machine charm exposes a cos-agent endpoint, and a machine grafana-agent charm collects the telemetry and forwards it to the COS applications (which typically run in a separate Kubernetes model).

• On Kubernetes, the opensearch-k8s charm exposes native COS endpoints (metrics-endpoint, grafana-dashboard, and logging), and a grafana-agent-k8s charm in the OpenSearch model collects the telemetry and forwards it to the COS applications. When COS runs in the same model, OpenSearch can integrate directly with the Prometheus, Grafana, and Loki applications instead.

Note

See: How to enable monitoring via COS and Grafana.

Additionally, Charmed OpenSearch can integrate with the OpenSearch Dashboards charm for exploring and visualizing your indexed business or application data. While COS with Grafana monitors your OpenSearch infrastructure and operational health, OpenSearch Dashboards provides purpose-built tools for interactive data exploration, custom visualizations, query builders, and OpenSearch-specific features like anomaly detection. Most production deployments benefit from using both: COS for infrastructure monitoring and OpenSearch Dashboards for data analysis.

Note

See: How to deploy OpenSearch Dashboards charm.

Metrics

The charm enables the Prometheus Exporter plugin for OpenSearch by default: The Prometheus Exporter Plugin for OpenSearch

The meaning of the metrics collected can be found in the upstream documentation:

• indices_stats_metrics

• nodes_stats_metrics

• cluster_stats_metrics

Alert rules

The charm deploys a pre-configured set of Prometheus alert rules by default.

To ensure you are referencing the latest default alert rules, check the source file of alert definitions in the repository:

• VM: machine prometheus_alerts.yaml

• K8s: kubernetes prometheus_alerts.yaml

The K8s charm additionally ships a set of Loki log-based alert rules, see opensearch.rules.

Logs

All the logs from the OpenSearch payload are available in the Grafana GUI at Home > Explore

To get OpenSearch logs, go to the Label filters field and set juju_application to the name of your OpenSearch application (for example, opensearch on VMs or opensearch-k8s on Kubernetes), select one operation, e.g. Line contains and run the query.

The following Grafana Explore screenshot shows OpenSearch logs from a VM deployment:

Grafana dashboard

The Charmed OpenSearch Grafana dashboard provides an at-a-glance view of cluster health, node resource utilization, and indexing throughput. The dashboard includes panels for CPU, memory, disk I/O, JVM heap, shard counts, and document indexing rates.

You can filter the displayed data using the selectors at the top of the dashboard:

• Juju model — the Juju model the cluster is deployed in

• Juju application — the Juju application (e.g. opensearch or opensearch-main)

• Juju unit — an individual Juju unit within the application

• Cluster — the OpenSearch cluster name (useful when multiple clusters share a COS instance)

• Role — filter by OpenSearch node role (e.g. cluster_manager, data)