Skip to main content

Health Checks

Overview​

Verify that OpsWorker is operating correctly across all components.

OpsWorker has no HTTP health endpoint and no liveness, readiness, or startup probes. You confirm health through three observable signals:

  1. Pod status: the agent pods are Running in the cluster
  2. Portal "Connected": the cluster shows Connected, driven by cluster-state messages the agent publishes to SQS
  3. Investigations completing: investigations finish and produce results

Pods in a Default Install​

A default install runs three pods in the opsworker namespace:

  • the OpsWorker agent
  • kubernetes-mcp-server
  • grafana-mcp

All three should be Running.

Verification Checklist​

Agent Health​

# Check pod status (expect three pods: agent, kubernetes-mcp-server, grafana-mcp)
kubectl get pods -n opsworker-agent

# Check agent logs for errors
kubectl logs -n opsworker-agent -l app.kubernetes.io/name=opsworker-agent --tail=50

There is no health probe to query. If the pods are Running and the portal shows Connected, the agent is healthy.

Portal Health​

  • Cluster status: Navigate to your cluster, which should show Connected. This state is driven by cluster-state messages the agent publishes to SQS.
  • Signal flow: Check Alerts for recent signals from your monitoring systems
  • Investigation flow: Check Investigations for recent completed investigations

End-to-End Test​

To verify the complete pipeline, trigger a real investigation and confirm each stage:

  1. An alert that matches a rule on the cluster arrives (or you start a manual investigation).
  2. The investigation runs against your cluster.
  3. Results appear in the portal under Investigations.
  4. A Slack notification is delivered (if Slack is configured).

Integration Health​

ComponentHow to Check
Agent connectivityCluster status in portal (connected when status is Success)
Alert ingestionNew signals appearing in the Alerts timeline
Investigation engineInvestigations completing successfully
SlackInvestigation notifications arriving in the mapped channel

Periodic Checks​

CheckFrequencyWhat to Look For
Agent pods runningDailyAll three pods are Running, no excessive restarts
Investigations completingDailyNo stuck investigations in "In Progress"
Slack deliveryWeeklyInvestigation notifications arrive in the mapped channel
Agent versionPeriodicallyAgent is on a recent version

Next Steps​