support network metrics for cgroupv2

[zouyee]

xref: #81

[akhilerm]

I am thinking if I can use this in containerd/containerd#10691 for getting the network metrics rather than via https://github.com/vishvananda/netlink that the CRI-O implementation uses for network metrics

cc: @mikebrow

[zouyee]

I am thinking if I can use this in containerd/containerd#10691 for getting the network metrics rather than via https://github.com/vishvananda/netlink that the CRI-O implementation uses for network metrics

cc: @mikebrow

https://github.com/opencontainers/runc/blob/main/libcontainer/container_linux.go#L152

[zouyee]

@dcantah @mikebrow PTAL

[mikebrow]

LGTM

[mikebrow]

I am thinking if I can use this in containerd/containerd#10691 for getting the network metrics rather than via https://github.com/vishvananda/netlink that the CRI-O implementation uses for network metrics

cc: @mikebrow

Will have to review the delta between using the NetworkStats returned from libcontainer through runc with the vishvananda/netlink code base.. probably do a review? @MikeZappa87 thoughts?

[MikeZappa87]

@mikebrow I am not sure how this repo fits together. Is this between the shim and the containerd?

[akhilerm]

@mikebrow I am not sure how this repo fits together. Is this between the shim and the containerd?

I think its more about checking if there is any difference between the metrics that is provided via netlink code which CRI-O uses for network metrics and the above implementation that we are using.

[mikebrow]

@mikebrow I am not sure how this repo fits together. Is this between the shim and the containerd?

Yes this is code/repo defines cgroups 1/2 and we use it in containerd when interacting with runtime engines through the shim. Eg. would be used in pulling in metric we are monitoring and conversion.. well it doesn't yet work for this new type but does for pid/cpu/memory... we're going to need to pipe the netstats through and thus begs the question which metrics are there differences.. needs investigation.