management cluster: Add missing rancher-agent airgap image

[hardys]

It has been reported that the rancher installation has failures when this image is not included, evidently it's not derived from the EIB inspection of the chart.

[hardys]

reported by @andreacasini

[andreacasini]

In my case this is what I see:

# k get po -n fleet-default
NAME                                                READY   STATUS             RESTARTS   AGE
rke2-machineconfig-cleanup-cronjob-29331365-69j4z   0/1     ImagePullBackOff   0          19m
rke2-machineconfig-cleanup-cronjob-29332805-jnpql   0/1     ImagePullBackOff   0          19m
rke2-machineconfig-cleanup-cronjob-29334245-jlltb   0/1     ImagePullBackOff   0          19m

Events:
  Type     Reason   Age                       From     Message
  ----     ------   ----                      ----     -------
  Normal   BackOff  3m28s (x14865 over 2d9h)  kubelet  Back-off pulling image "registry.rancher.com/rancher/rancher-agent:v2.11.3"
  Warning  Failed   3m14s (x14866 over 2d9h)  kubelet  Error: ImagePullBackOff

# /var/lib/rancher/rke2/bin/crictl images | grep agent
registry.rancher.com/rancher/fleet-agent                           v0.12.4

@hardys @alknopfler carefult with the version of the agent in the commit because in my case for example I'm on edge 3.3 and I'm using rancher v2.11.3 whle the PR is pointing to v.2.12.1.

I assume it needs to be backported with proper versioning for all editions.

[e-minguez]

Just in case, it is here https://prime.ribs.rancher.io/rancher/v2.12.2/rancher-images-digests-linux.txt

[hardys]

Just in case, it is here https://prime.ribs.rancher.io/rancher/v2.12.2/rancher-images-digests-linux.txt

Interestingly it seems the shell version we're specifying is wrong, confirmed via helm template too:

helm template rancher-prime/rancher --version 2.12.1 | grep image:
      - image: registry.rancher.com/rancher/rancher:v2.12.1
          image: "rancher/shell:v0.5.0"
          image: "rancher/shell:v0.5.0"

Perhaps we can remove that, since unlike the agent image which doesn't appear in the template output, IIUC EIB should derive that one correctly from the chart?

[hardys]

I just updated the shell version for now, perhaps we can test removing it in future