Cluster Health Analyzer

An analyzer for OpenShift cluster health data.

Overview

The Cluster Health Analyzer process incoming stream of health signal from the OpenShift cluster and enriches it to provide better views at the data to enhance the troubleshooting experience.

It provides:

• incidents detection: heuristics to group individual alerts together to allow better reasoning about the root cause of the issues.
• components mapping and ranking: opinionated way to assign the alerts to high-level components and ranking based on the importance of the components from overall cluster health perspective.

Install

Login to a cluster using oc login command.

oc apply -f manifests/backend -f manifests/frontend

Usage

Cluster health analyzer is a backend that exposes the results via Prometheus metrics:

# Mapping of source signal to components and incident groups.
cluster:health:components:map
{
  # Identifier of the source signal type
  type="alert"

  # Matchers against the source labels
  src_alertname="KubeNodeNotReady",
  src_namespace="openshift-monitoring",
  src_severity="warning",

  # Identifier of the mapped component.
  layer="compute",
  component="compute",

  # Incident group id
  group_id="b8d9df3f-8245-4f5a-825d-15578a6c8397",

# Value represents the impact on the component severity
} -> 1

# Metadata about the components in the system
cluster:health:components
{
  # Identifier of the component
  component="compute", layer="compute"

# The value represents the ranking of the component: the lower number the higher
# importance of the component.
} -> 1

See https://github.com/openshift/cluster-health-console-prototype of an example usage of the data for incidents navigation.

Development and testing

If you want to contribute to the project head over to development.md