k8s: guest-pull: Kill all processes in container test fails when pulling the image inside the guest

[fidencio]

# Events:
#   Type     Reason     Age                From               Message
#   ----     ------     ----               ----               -------
#   Normal   Scheduled  90s                default-scheduler  Successfully assigned kata-containers-k8s-tests/busybox to 984fee00bd70.jf.intel.com
#   Normal   Pulling    82s                kubelet            Pulling image "quay.io/prometheus/busybox:latest"
#   Normal   Pulled     82s                kubelet            Successfully pulled image "quay.io/prometheus/busybox:latest" in 373ms (373ms including waiting)
#   Normal   Created    82s                kubelet            Created container first
#   Normal   Started    80s                kubelet            Started container first
#   Normal   Pulled     79s                kubelet            Successfully pulled image "quay.io/prometheus/busybox:latest" in 420ms (420ms including waiting)
#   Warning  Failed     78s                kubelet            Error: failed to create containerd task: failed to create shim task: the file sleep was not found: unknown
#   Warning  Failed     76s                kubelet            Error: failed to create containerd task: failed to create shim task: failed to mount /run/kata-containers/2ccc250bd74a81ef8c525b6acecaf3ff9c00cc8a2700c3ef17446c75374e32d7/rootfs to /run/kata-containers/first-test-container/rootfs, with error: ENOENT: No such file or directory: unknown
#   Normal   Pulled     59s (x2 over 77s)  kubelet            Successfully pulled image "quay.io/prometheus/busybox:latest" in 364ms (364ms including waiting)
#   Normal   Pulling    29s (x4 over 80s)  kubelet            Pulling image "quay.io/prometheus/busybox:latest"
#   Normal   Pulled     29s                kubelet            Successfully pulled image "quay.io/prometheus/busybox:latest" in 357ms (357ms including waiting)
#   Normal   Created    29s (x4 over 79s)  kubelet            Created container first-test-container
#   Warning  Failed     29s (x2 over 59s)  kubelet            Error: failed to create containerd task: failed to create shim task: load image bundle

And then taking a look at the test itself, we see:

setup() {
...
yaml_file="${pod_config_dir}/initcontainer-shareprocesspid.yaml"
...
}

...
@test "Kill all processes in container" {
        # Create the pod
        kubectl create -f "${yaml_file}"
...
}

And when taking a look at the error, it makes me think that nydus is not properly being used for initContainers.

@ChengyuZhu6, mind to verify if it works on your end?

[wainersm]

Hi @fidencio ! One question: did you get this failure when setting shared_fs=none and the runtime handler annotation?

Asking because I'm getting this error even when shared_fs=9p (i.e. using 9p) too. I realized that because I'm run a set of tests for qemu-coco-dev with setting the runtime handler annotation ("io.containerd.cri.runtime-handler").

Another test that's failing and I haven't seen you report is k8s-empty-dirs.bats . Wondering why.