containernetworking / plugins

Some reference and example networking plugins, maintained by the CNI team.
Apache License 2.0
2.23k stars 788 forks source link

host-device continuously outputs 'Link not found' error, and Pod will stuck on Terminating state #995

Closed qmloong closed 7 months ago

qmloong commented 11 months ago

I am using multus-cni, which defaults to using host-device to mount RDMA network card devices. However, some exceptional scenarios have arisen, causing an error when kubelet executes the RunPodSandBox interface: 'failed to find device name for pci address'. When kubelet automatically performs sandbox cleanup, it continuously outputs the 'Link not found' error in the netns of the erroneous sandbox, as it cannot find the corresponding network card device. This, in turn, causes the pod to be blocked in the Terminating state.