Skip to main content
Start an investigation from an open incident when you need a cited explanation across the systems connected to OpenCluster.

​
Prerequisites

  • The deployment has a configured and consented model provider.
  • The incident is open.
  • At least one relevant integration is connected and verified.
  • You have the Editor or Admin role.
Alertmanager supplies the incident context. Slack and GitHub are the live sources an investigation can read. Kubernetes supplies a current workload inventory for orientation.

​
Start the investigation

Open Incidents, select the incident, and start an investigation. You can also ask OpenCluster to investigate in plain language. If the question matches none or several open incidents, OpenCluster asks for clarification instead of choosing one. The investigation continues in the background. A deployment runs a bounded number at once; if capacity is full, the request is refused and no partial investigation is created.

​
Review the result

Check the outcome first: Then review findings in causal order, open their cited reads, and check whether any result was truncated or failed. Treat next steps as recommendations for an operator to evaluate; OpenCluster does not execute them. See Understand findings and citations for the meaning of finding types, confidence, and citations.