Back Matter

Troubleshooting Index

SymptomLikely CauseFix
SSH to Spark 2 times outiptables blocked after k3s restartFlush iptables — see Ch 2
DNS not resolvingsystemd-resolved cache stalesudo resolvectl flush-caches
GPU not visible in k8sGPU Operator pods not RunningCheck kubectl get pods -n gpu-operator
Ray shows GPU: 1 not 2Worker pod not connectedCheck worker logs; verify DNS address
"exec format error"x86 image on ARM64Use nvcr.io/nvidia/vllm:25.09-py3
Model download failsHF token missing/invalidCheck kubectl get secret hf-token -n core-services
AIBrix CrashLoopBackOffDependencies not installed firstApply aibrix-dependency before aibrix-core
NGC image pull failscontainerd lacks NGC credentialsCreate docker-registry Kubernetes secret — see Ch 5