HAMi
HAMi copied to clipboard
if use gpu-operator install nvidia driver, when node restart, driver-plugin start CrashLoopBackOff
- when the node reboots after,
gpu-operator
can reinstall NVIDIAdriver
; - but the same time
device-plugin
pod starts, but thisdriver
does not install completely. so pod in CrashLoopBackOff.
we can add a initContainers
to check nvidia driver to resolve this problme.