Skip to main content
Version: Next

Observability

OpenChoreo provides an optional observability plane, which consists of a comprehensive observability stack for monitoring applications deployed on the platform. The observability plane is designed in a modular architecture so platform engineers can configure the observability plane to fit their needs, using the tools and technologies they are familiar with.

The observability plane ships a selected set of modules by default when installed, but platform engineers can replace or supplement the default modules with any community module.

Overview​

OpenChoreo's default observability architecture consists of:

PillarComponents
LogsA logs collector deployed to the data plane, and a logs storage deployed to the observability plane (defaults to observability-logs-opensearch module)
MetricsA metrics collector deployed to the data plane, and a metrics storage deployed to the observability plane (defaults to observability-metrics-prometheus module)
TracesA traces collector deployed to the data plane, and a traces storage deployed to the observability plane (defaults to observability-tracing-opensearch module)

All observability data is accessible through the Observer API, which provides a unified interface for querying logs, metrics, and traces.

Alerting is a feature provided by the observability plane, which is integrated with logs and metrics modules to facilitate a unified alerting experience across the platform.

Architecture​

Single-Cluster Setup​

In single-cluster mode, all planes run in the same Kubernetes cluster. Observability data is collected directly from the data planes and workflow plane via the agents deployed in the observability plane.

Multi-Cluster Setup​

Multi-cluster mode is designed to run the observability plane on a dedicated cluster. Data planes and workflow plane deploy local collectors from the respective observability modules that publish observability data to the observability plane through the gateway ingress of the observability plane.

In this setup:

  • Data Plane deploys Fluent Bit to collect logs, Prometheus Agent to collect metrics, and OpenTelemetry Collector to collect traces
  • Workflow Plane deploys Fluent Bit to collect workflow run logs
  • All collectors publish data through the Gateway Ingress in the Observability Plane
  • The Observer API queries OpenSearch and Prometheus to serve unified observability data

For detailed multi-cluster setup instructions, see Multi-Cluster Connectivity.

Prerequisites​

  • OpenChoreo control plane installed
  • Data plane and workflow plane (optional) installed to observe
  • Observability plane installed (see Installation)

Installing the Observability Plane​

Refer to Getting Started or Multi-Cluster Connectivity for instructions on installing the observability plane in single-cluster or multi-cluster mode.

In This Section​

Configuration Reference​

Key Helm Values​

ValueDefaultDescription
observability-logs-opensearch.enabledfalseEnable OpenSearch based community module for logs
observability-metrics-prometheus.enabledfalseEnable Prometheus based community module for metrics
observability-tracing-opensearch.enabledfalseEnable OpenSearch based community module for traces
rca.enabledfalseEnable AI SRE Agent

For complete configuration options, see the Observability Plane Helm Reference.