Description

container exec may hang on containerd version 1.7.14. We believe that this problem was caused by #9928 with a possible fix coming via #9999.

Steps to reproduce the issue

Using Kubernetes version 1.26 or later (likely any version), run the following loop multiple times against the same pod. One or more of the exec commands will hang indefinitely within a few minutes.

while true; do
    kubectl --request-timeout 30s exec -n "${namespace}" "${pod_name}" -- "${command}" >/dev/null
done

Describe the results you received and expected

Kubernetes container exec probes and kubectl exec commands should not hang but may.

What version of containerd are you using?

1.7.14

Any other relevant information

No response

Show configuration if it is related to CRI plugin.

No response