Use dl.k8s.io for getting kubectl

[superseb]

Issue: #1260

Problem

Preparing for GCS bucket deprecation at some point (kubernetes/k8s.io#2396)

Solution

Change to the currently documented source

Testing

Using start_testacc.sh

Engineering Testing

Manual Testing

Automated Testing

QA Testing Considerations

Regressions Considerations

[a-blender]

@superseb Hey is this ready for review? Is there a linked issue? Any thoughts on the scope of testing this issue... I'm not entirely sure when/how the TF provider uses kubectl. It uses a token to access the Rancher API client and accesses downstream resources that way.

[superseb]

According to the script it is being used here https://github.com/rancher/terraform-provider-rancher2/blob/master/scripts/start_testacc.sh#L185.

I had two Drone builds being killed so I left it but will create the corresponding issue but this is the same as the one for terraform-provider-rke.

[a-blender]

Can you verify make testacc still passes? Otherwise lgtm

[superseb]

@a-blender There seems to be no reference to this file anywhere, can we just remove it instead?

[a-blender]

@superseb Not entirely sure, running make testacc runs the infrastructure tests but right now it continuously tries to create a k3s cluster without being fully successful .

+++ /Users/annablendermann/.rd/bin/docker run -i --rm curlimages/curl -sk https://172.17.0.2:6443/ping
+++ echo starting rancher k3s cluster
++ k3s_ready='starting rancher k3s cluster'
++ '[' 'starting rancher k3s cluster' '!=' pong ']'
++ sleep 5

I think tf will need KUBECTL_URL to download the kubectl binary to access the cluster if it doesn't already exist. I'd probably just update it for now.

Is that what you were referring to?

[superseb]

My point is that scripts/start_testacc.sh is not referenced anymore in the code. The reference to it was removed a long time ago here.

Unless there is an external dependency that uses is, or if there is a case to keep it, I would remove it to clean the codebase from unused scripts.