[BUG] k3d provider can't join more than one worker node to a cluster with only a master node

[sunyakun]

Describe the bug

k3d 集群初次创建时如果只有一个 master 节点，那么使用 autok3s join 命令则最多只能添加一个 worker 节点，当超过一个 worker 节点后再次执行 autok3s join 则会失败

To reproduce

1. 使用 autok3s 创建一个只有一个 master 节点的 k3d 集群

   $ autok3s create -p k3d --name test --master 1

2. 向 k3d 集群添加一个 worker 节点

   $ autok3s join -p k3d --name test --worker 1

3. 继续向集群添加 worker 节点

   $ autok3s join -p k3d --name test --worker 1

Expected behavior

期望命令全部执行成功，最终 k3d 集群有一个 master 节点和两个 worker 节点

Screenshots

Environments (please complete the following information):

• OS: docker image cnrancher/autok3s:v0.8.0
• AutoK3s Version {"gitVersion":"v0.8.0","gitCommit":"e4f13c2fb22e6b12859cc9d5f21501409927803c","gitTreeState":"","buildDate":"2023-05-15T08:07:40Z","goVersion":"go1.19.9","compiler":"gc","platform":"linux/amd64"}

[JacieChao]

Thanks for the feedback. When adding the agent node one by one, the K3d agent name is duplicated when stitching the K3d Node name.

As a workaround, please add or join multiple worker nodes at one time. We will fix this issue in the next version.

[sunyakun]

By the way, if the join command fails to execute, autok3s will attempt to perform a rollback, including deleting the kubectl context config. I believe this behavior is unexpected.

[JacieChao]

Yes, rollback with failure won't remove the kubectl context config when joining nodes. Will fix this issue too.

[sunyakun]

I would be happy to submit a PR to fix this bug. You can assign this issue to me.

[JacieChao]

Sure. Thanks