Added log collection support for vanilla Kubeadm clusters

[gnuninu]

This PR addresses issue/request #209

New functions created:

    kubeadm-k8s
    kubeadm-certs
    kubeadm-etcd

the switch -r now also includes kubeadm as a selector from the list

[leodotcloud]

It's possible that a user doesn't store the kubeconfig in this location. Can we make this configurable?

KUBECONFIG=${KUBECONFIG:-"$USER/.kube/config"}

[leodotcloud]

Make kubectl binary location also configurable.

KUBECTL_CMD=${KUBECTL_CMD:-"/usr/bin/kubectl"}

[leodotcloud]

Curious, why do we have two "if" blocks?

[leodotcloud]

I think these if checks can be done once at the top of the function, not needed for every section.

[leodotcloud]

Add ETCD_CMD checks. This binary might not exist on the host.

[gnuninu]

thanks @leodotcloud for reviewing and for the good suggestions.
My attempt was to reuse parts of the already existent rancher code and not altering things much by keeping the same style/structure.
This PR it's just an attempt to introduce a "fallback" collection that should be sufficient for the most common scenario described in the k8s docs (kubeadm and kubectl binary location, etcd dirs and so on...).
No doubt it can be improved with the help of the wider community to support more environments and i would be willing to help by sending future PRs.
However, may i suggest to accept the code as it is for the time being so at least a log collection from kubeadm is in place? i have tested from a couple of brand new test installs i have and it works.
This PR leaves rancher collection untouched and everything will continue to work as expected.
Perhaps can we create a new issue using your points as a basis for an improvement, hence a new PR?

[leodotcloud]

I appreciate your PR and am glad you could reuse some parts, but the existing code makes some assumptions and those are not applicable to your changes.

may i suggest to accept the code as it is for the time being

That's not how code reviews work anywhere :)

Anyways, I quickly made the changes that I suggested.

[gnuninu]

Anyways, I quickly made the changes that I suggested.

Thanks appreciated :)

may i suggest to accept the code as it is for the time being

That's not how code reviews work anywhere :)

I welcome your criticism but have you actually tested your modification ? :)
I have tried to fetch it and run it from my kubeadm env, looks like the automatic detection is not working.

wget https://raw.githubusercontent.com/rancherlabs/support-tools/master/collection/rancher/v2.x/logs-collector/rancher2_logs_collector.sh

sudo ./rancher2_logs_collector.sh 
2023-02-01 15:30:33: Created ./tmp.jchaD0DGbD
2023-02-01 15:30:33: Detecting available commands... renice ionoice
2023-02-01 15:30:33: Detecting OS... opensuse-leap 15.4
2023-02-01 15:30:33: Detecting k8s distribution... 
2023-02-01 15:30:33: couldn't detect k8s distro

It's missing this part in the sherlock function from my original PR:

      if $(command -v kubeadm >/dev/null 2>&1)
        then
          if $(kubeadm version >/dev/null 2>&1)
            then
              DISTRO=kubeadm
              echo "kubeadm"
            else
              FOUND+="kubeadm"
          fi
      fi

I am more than happy to send a new PR to fix this

[leodotcloud]

1. Yes, I tested the script.
2. I removed that piece of logic. We don't want any unintended side effects as the new code has not been tested thoroughly by our team.

For your use-case, you have to invoke the script with -r kubeadm.

[gnuninu]

2. I removed that piece of logic. We don't want any unintended side effects as the new code has not been tested thoroughly by our team.

An automatic detection would be more appropriate for my case because the idea was to embed this script into other projects where we have mixed environments, Rancher and Kubeadm.
This is what led me to create this PR in the first place.