We noticed e2e test failures on Amazon Linux ECS Optimized OS. It runs with cgroupv1. The kind nodes are using docker and the pods use containerd. All the pods are running containers in the same cgroup namespace but this ns is the one allocated to the docker nodes, not the host cgroup namespace. Thus, we don't verify that the cid is found in /proc/self/cgroup whereas we should. We also don't send the cgroup inode.
We noticed e2e test failures on Amazon Linux ECS Optimized OS. It runs with cgroupv1. The
kindnodes are using docker and the pods use containerd. All the pods are running containers in the same cgroup namespace but this ns is the one allocated to the docker nodes, not the host cgroup namespace. Thus, we don't verify that the cid is found in /proc/self/cgroup whereas we should. We also don't send the cgroup inode.