Troubleshooting

Start from the symptom. Every entry links to the page with the full runbook.

Twenty-odd pages carry a Troubleshooting section, which is only useful if you already know which component is at fault. Start here instead — find the symptom, follow the link.

Something is unreachable

SymptomLikely causeWhere to look
504 Gateway Timeout on a public URLTraefik cannot reach the cluster — NetBird tunnel or routeNetBird Peer
Every LoadBalancer IP goes dark after a Cilium restartStale L2 announcement lease; the holder renews but stops answering ARPUpgrade Guide, then just gateway-repair-l2
A *.madhan.app host resolves but nothing answersService is LAN-only and you are off the LAN and off the VPNService Access
TLS certificate errors on a serviceWildcard cert or the Cilium secret mirrorcert-manager
Argo CD UI loads but lists no applicationsUsually a stale aggregated APIService breaking API discovery — check kubectl api-resources for errorsArgo CD

A pod will not start

SymptomLikely causeWhere to look
GPU pod stays Pending with no clear reasonMissing the dedicated=ai toleration — worker4 is taintedGPU
GPU pod runs but CUDA is invisibleMissing runtimeClassName: nvidiaGPU
Multi-Attach error for volumeRWO volume still held by the old pod during a rolloutHarbor
PVC stuck Pending, pod stuck ContainerCreatingLonghorn volume cannot attachLonghorn
No Longhorn volume on a node will attach after a Talos upgradeNew open-iscsi rejects a record its predecessor wrotejust longhorn-repair-iscsiUpgrade Guide
Pod cannot find its k8s SecretNothing mounts the CSI volume, so the sync never firesSecrets
ComfyUI returns 503It ships scaled to zerojust comfyui onComfyUI

Something is missing or wrong

SymptomLikely causeWhere to look
kubectl top says Metrics API not availableMetrics ServerMetrics Server
Metrics missing from GrafanaVMAgent is not scraping the targetVictoriaMetrics
Ollama cannot load a modelComfyUI is holding VRAM, and the error surfaces on OllamaGPU
Grafana OIDC login loops or errorsroot_url / redirect URI mismatchGrafana
Falco UI never becomes readyRedis wedged with MISCONF — TTL and volume sizeFalco
App stuck OutOfSync in Argo CDSync wave, hook failure, or SSA ownership conflictArgo CD
A manual kubectl change reverted itselfselfHeal=true — this is intendedArgo CD

A node is in trouble

SymptomLikely causeWhere to look
Node will not reboot after a Cilium problemUnmount blocks on an NFS share-manager ClusterIP it can no longer reachUpgrade Guide — powercycle
Talos upgrade hangs draining a nodeLonghorn and single-instance CNPG PDBs allow zero disruptionsjust talos-upgrade <ip> <schematic> false
Cluster health check fails between node upgradesStop and fix before continuingjust talos-health

First commands

just talos-health                       # cluster-level health
kubectl get applications -n argocd      # what Argo CD thinks is wrong
kubectl api-resources 2>&1 >/dev/null   # any output here means broken discovery
kubectl get events -A --sort-by=.lastTimestamp | tail -30