Skip to main content
This path creates an incident from Alertmanager and gives its investigation a Slack conversation to inspect.

​
Before you start

You need:
  • a running OpenCluster deployment with a model provider configured;
  • an OpenCluster account with the Admin role;
  • Prometheus Alertmanager 0.29.0 or later;
  • permission to create and install a Slack app.
If you operate OpenCluster, complete the self-hosted setup first.

​
1. Connect Alertmanager

In Integrations, choose Prometheus Alertmanager, enter a name, and create the integration. Copy the webhook URL and secret from the response. The secret is shown once. Add the receiver to Alertmanager and route a test alert to it:
Reload Alertmanager. See Prometheus Alertmanager for routing options and delivery responses.

​
2. Connect Slack

Create a Slack app with these bot token scopes:
  • channels:read
  • channels:history
  • users:read
Install the app, copy its xoxb- token, and paste the token into a new Slack integration. Invite the app to a public channel used for incident response. Open the integration and select Verify. A bot token normally verifies as degraded because Slack does not grant bots search:read; channel and thread reads are still available. See Slack for the full access model.

​
3. Deliver an alert

Fire a test alert. Add a short message about the alert to the Slack channel, including the affected service or alert name so the conversation is useful during investigation. Verify the Alertmanager integration. It should report the time of the accepted delivery. Open Incidents and select the incident created from the alert.

​
4. Run the investigation

From the open incident, start an investigation. An Editor or Admin can start one. It runs in the background. The quickstart succeeds when the finished record shows:
  • status: concluded with no stoppedBy value;
  • Slack among the sources made available for the incident;
  • at least one recorded Slack read and its result;
  • cited findings when the recorded reads support a conclusion;
  • unresolved questions when the cause cannot be established.
An investigation may conclude with no findings and still verify this path. That means the Slack read completed, but the available data did not support a defensible cause. A concluded record with stoppedBy is a partial conclusion forced by a configured limit; a failed record produced no valid conclusion. Neither verifies the quickstart. Use Troubleshooting before trying again. Next, read How investigations work or add GitHub for deployment and code context.