[release/1.6 backport] moving up to go-cni v1.1.4#6728
Conversation
Signed-off-by: Mike Brown <[email protected]> (cherry picked from commit a51da91) Signed-off-by: Sebastiaan van Stijn <[email protected]>
|
Hmm.... rate limits? Are we missing IP addresses again on Hub? Let me check internally vagrant ssh -- curl https://api64.ipify.org/
199.19.85.81 |
|
@mikebrow I added |
Nod. The config change is in code vs cni config in storage. You get loopback unless it's switched off |
|
@mikebrow I think most of people might just install the eth0's CNI, like flannel 😂 . The CRI will load the loopback with https://github.com/containerd/containerd/blob/main/pkg/cri/server/service_linux.go#L95. The following bash result is from my local. |
|
Yes. By "You get loopback unless it's switched off" I mean unless networking is switched off for the pod or the networking config path is set to null. Defaults with the default CNI plugin config (before this pr noting 0.3.1 for lo): |
|
understand. I just concerned that if the cluster is still using v0.3.1 loopback binary, the cni will not work after upgrade to containerd with this commit. I was facing this issue like 'loopback only supports 0.3.1, not v1.0.0' in my local. I have to upgrade the loopback binary version. So I am ok with this commit but still want to put changelog label here because the new 1.6 minor release needs upgrade loopback binary. |
ah.. ok.. to many versions cni lib, cni config(s), plugin binaries... |
backport of #6721 (re-ran go mod tidy and vendor)
(cherry picked from commit a51da91)