The right kubectl is missing on old kubernetes [rc1]

[kravciak]

Job kubewarden-controller-delete-controller-job fails with kubectl missing error

To reproduce with e2e tests:
make clean cluster install opentelemetry-remote-collector-tests.bats K3S=1.25 VERSION=next

Error when running e2e

~ k get pods -n kubewarden
my-collector-collector-7d4986f75-9ktbb              1/1     Running   0             2m4s
kubewarden-controller-759f6d4847-sgqvd              1/1     Running   0             2m1s
policy-server-default-75479dbb6-pf54m               1/1     Running   0             94s
kubewarden-controller-delete-controller-job-fwmj7   0/1     Error     4 (57s ago)   109s

~ k logs  -n kubewarden kubewarden-controller-delete-controller-job-fwmj7
F0122 15:49:36.491508       1 main.go:81] the right kubectl is missing, binary downloads from kubernetes' upstream mirror are disabled

I also noticed we pull kuberlr-kubectl:v4.0.0 which is meant for rancher v2.11.0+
https://github.com/rancher/kuberlr-kubectl?tab=readme-ov-file#branches-releases-and-rancher

Failed release test: https://github.com/kubewarden/helm-charts/actions/runs/12910158312/job/36003339620

[kravciak]

Seems it's caused by AllowDownload = false in their Dockerfile

Can be tested also with: make clean cluster install uninstall K3S=1.25 VERSION=next
Affected versions: k8s 1.25 - 1.27

[viccuad]

Opened rancher/kuberlr-kubectl#42.

[viccuad]

Blocked on that PR; if it doesn't move forward I will either fork or go back to our image for this release at least.

[flavio]

The bug has been fixed by doing this revert. We can close the issue

[viccuad]

The follow-up of this is flavio/kuberlr#153.