getcwd error in thick plugin

[bc185174]

What happend:

No pods are being assigned IPs. Multus is frequently logging:

2024-07-11 14:24:14.962 [INFO][192098] k8s.go 598: Teardown processing complete. ContainerID="c011cf5a6a0cfe50c6061b3373844c9d65bc1a31e664677f99d61a93be8b0e03"
shell-init: error retrieving current directory: getcwd: cannot access parent directories: No such file or directory

What you expected to happen:

No error and pod to be assigned IP.

How to reproduce it (as minimally and precisely as possible):

Deploy latest version multus from master branch

Anything else we need to know?:

Tried using v4.0.2 but get service account errors due to the exec running chroot on host (should really avoid doing this, its a security nightmare).

Environment:

• Multus version: master image path and image ID (from 'docker images')
• Kubernetes version (use kubectl version): 1.28.9
• Primary CNI for Kubernetes cluster: Calico
• OS (e.g. from /etc/os-release): Ubuntu
• File of '/etc/cni/net.d/': N/A
• File of '/etc/cni/multus/net.d': N/A
• NetworkAttachment info (use kubectl get net-attach-def -o yaml): N/A
• Target pod yaml info (with annotation, use kubectl get pod <podname> -o yaml): N/A
• Other log outputs (if you use multus logging): N/A

[dougbtv]

Tried using v4.0.2 but get service account errors due to the exec running chroot on host (should really avoid doing this, its a security nightmare).

This is essentially required for a CNI thick plugin, as it's a replacement for running directly on the host -- where you do have full access to the host

[dougbtv]

I think we need to pod spec and net-attach-def to replicate it? Otherwise, I'd think we'd hit it in our e2e tests, we'd love a reproducer, thanks