Ihor Indyk

We are experiencing the same issue on AKS 1.30.3 running linkerd stable-2.14.9. The linkerd-cni tokens have started to expire every hour resulting new pods are not able to start with...

A pull request for the possible fix was raised https://github.com/linkerd/linkerd2-proxy-init/pull/416

@alpeb This is what happens when token is being rotated ``` CREATE:ISDIR ..2024_09_05_17_44_30.3462809433 ATTRIB:ISDIR ..2024_09_05_17_44_30.3462809433 CREATE ..data_tmp MOVED_FROM ..data_tmp MOVED_TO ..data ``` At this point you would have new token....

@alpeb here's a piece of sidecar container code we use in linkerd CNI pods as a workaround. The line with `sed` is just a trick to change the CNI config...

@alpeb sure no problem Just deployed it to a smaller non-prod AKS cluster with 13 nodes I'll monitor it if everything is grand will deploy it to 2 bigger non-prod...

It didn't go well though On a first node the plugin was successfully replaced by the new version On the second it has stuck with this error: ``` Events: Type...

@alpeb here you go ``` k logs linkerd-cni-kv95k -n linkerd-cni --context lz-nonprod-sa-aks Defaulted container "install-cni" out of: install-cni, repair-controller Error from server (BadRequest): container "install-cni" in pod "linkerd-cni-kv95k" is waiting...

As I mentioned above after I drained and deleted the node where linkerd CNI could not start the cluster added a new node where the plugin has started with no...