inspect-cluster

name	inspect-cluster
metadata	{"author":"rhoai3-coding-demo","version":"1.0.0","platform-family":"rhoai","platform-baseline":"repo","ocp-baseline":"repo","skill-group":"Demo Environment"}
description	Safely gather OpenShift cluster state for the RHOAI demo. Use when troubleshooting deployment failures, validating stage completion, or gathering diagnostic information. Must be used in readonly mode — no mutations. Do NOT use for remediation (use rhoai-troubleshoot), resource scaling (use manage-resources), or manifest review (use review-manifest-compliance or review-doc-alignment).

Inspect Cluster

Gather diagnostic information from the cluster without modifying anything. Summarize findings so the caller can decide on actions.

When to invoke

• Troubleshooting a deployment failure
• Validating stage completion
• Gathering pre-change cluster state
• Diagnosing pod or operator issues

Prerequisite

Verify oc login status before running any commands:

oc whoami --show-server

If not logged in, report that immediately instead of failing on every command.

Standard inspection sequence

When asked to inspect a stage or component:

1. Check ArgoCD Application sync status:

oc get application <stage-app-name> -n openshift-gitops \
  -o jsonpath='{.status.sync.status}/{.status.health.status}'

2. Check pod status in the target namespace:

oc get pods -n <namespace> -l app.kubernetes.io/part-of=<component>

3. For failing pods, get events and recent logs:

oc describe pod <pod-name> -n <namespace> | tail -30
oc logs <pod-name> -n <namespace> --tail=50

4. For InferenceServices, check readiness:

oc get isvc -n <namespace>

5. For operators, check CSV status:

oc get csv -n <namespace> | grep -i <operator>

Key namespaces

Namespace	Components
redhat-ods-applications	RHOAI Dashboard, DSC, GenAI Studio
redhat-ods-operator	RHOAI Operator
openshift-gitops	ArgoCD
openshift-operators	Subscriptions (NFD, GPU, Service Mesh)

Additional namespaces are created as stages are deployed. Check Argo CD Applications for the target namespace of each stage.

Output format

Return a structured summary:

Component: <name>
ArgoCD: <Synced/OutOfSync> / <Healthy/Degraded>
Pods: <X/Y ready>
Issues: <list of problems found, or "None">
Recommendation: <what to do next>

Safety constraints

• Never run oc delete, oc patch, oc scale, or oc apply
• If a stage has a validate.sh script, suggest running it for comprehensive checks

Related skills

• rhoai-troubleshoot — active troubleshooting and remediation
• manage-resources — scale or modify cluster resources
• validate-demo-step — structured stage validation