GKE Platform Troubleshooting

Decision-tree playbooks for the most common failure modes encountered in this cluster.

When to Use

• An Argo CD application shows Unknown, Degraded, or OutOfSync and is not self-healing.
• Pods in the platform namespace are in CrashLoopBackOff, Pending, or Error state.
• PVCs are stuck in Pending.
• TiKV, PD, or SurrealDB won't start.
• GCP quota errors appear in events.

Playbook 1: Argo CD application stuck at Unknown

Diagnose:

kubectl describe application <APP_NAME> -n argocd | grep -A 5 "Conditions"