Need Allow Specific INSTALL_K3S_VERSION

autok3s <sub-command> \
--k3s-version

• Add Command Return Supported K3s Version.
• UI better use drop-down component show which K3s version can be selected.
  e.g.

autok3s k3s-versions

Use the --master-extra-args '--docker' & --worker-extra-args '--docker' parameter to start the autok3s cluster. The cluster can be created successfully, but the startup fails:

Sep 25 11:37:25 iZ2ze75abrpw8ky4hu2onxZ k3s[7684]: I0925 11:37:25.692147    7684 client.go:75] Connecting to docker on unix:///var/run/docker.sock
Sep 25 11:37:25 iZ2ze75abrpw8ky4hu2onxZ k3s[7684]: I0925 11:37:25.692351    7684 client.go:104] Start docker client with request timeout=2m0s
Sep 25 11:37:25 iZ2ze75abrpw8ky4hu2onxZ k3s[7684]: F0925 11:37:25.692740    7684 server.go:269] failed to run Kubelet: failed to create kubelet: failed to get docker version: Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?

Using curl https://get.docker.com | VERSION=19.03 sh -s - --mirror Aliyun directly to install docker on the Alibaba Cloud host also fails. It is recommended to confirm the status of docker before starting K3S

Need to support Aliyun(implements aliyun cloud provider):

APIs

• ecs.RunInstances
• ecs.DescribeInstances
• ecs.DescribeInstanceStatus
• ecs.DeleteInstances
• ecs.DescribeVSwitches
• ecs.DescribeVpcs
• ecs.ListTagResources
• ecs.StartInstances
• ecs.StopInstances
• vpc.TagResources
• vpc.ListTagResources
• vpc.AllocateEipAddress
• vpc.ReleaseEipAddress
• vpc.AssociateEipAddress
• vpc.UnassociateEipAddress

RAMs

{
    "Statement": [
        {
            "Action": [
                "ecs:CreateNetworkInterface",
                "ecs:DescribeNetworkInterfaces",
                "ecs:AttachNetworkInterface",
                "ecs:DetachNetworkInterface",
                "ecs:DeleteNetworkInterface",
                "ecs:DescribeInstanceAttribute",
                "ecs:DescribeInstanceTypesNew",
                "ecs:AssignPrivateIpAddresses",
                "ecs:UnassignPrivateIpAddresses",
                "ecs:DescribeInstances",
                "ecs:DeleteInstances",
                "ecs:DescribeInstanceStatus",
                "ecs:RunInstances",
                "ecs:ListTagResources",
                "ecs:StartInstances",
                "ecs:StopInstances"
            ],
            "Resource": [
                "*"
            ],
            "Effect": "Allow"
        },
        {
            "Action": [
                "vpc:DescribeVSwitches",
                "vpc:CreateVSwitch",
                "vpc:DeleteVSwitch",
                "vpc:DescribeVSwitches",
                "vpc:DescribeVpcs",
                "vpc:TagResources",
                "vpc:AllocateEipAddress",
                "vpc:AssociateEipAddress",
                "vpc:DescribeEipAddresses",
                "vpc:UnassociateEipAddress",
                "vpc:ReleaseEipAddress",
                "vpc:ListTagResources"
            ],
            "Resource": [
                "*"
            ],
            "Effect": "Allow"
        }
    ],
    "Version": "1"
}

Components

• AppHub
• Terway
• Cloud-Controller-Manager(Alibaba)

We should disable traefik by default when enabling ccm.

Need to add CI Process

Encrypt & Store Sensitive Data:

1. credential data like (access-key/access-secret)
2. cluster sensitive data like (cluster token/access-key/access-secret)

Need to support Aliyun Terway CNI plugin

• Terway ENI mode

Enhancement SSH Tunnel:

• password login
• passphrase key-pair login
• ssh-agent support

The current naming scheme does not satisfy certain scenarios and is vulnerable to intrusion. Need to use Tag instead.

InstanceName: e.g.

autok3s.<cluster name>.<master>/<worker>

InstanceTags: e.g.

autok3s: true
cluster: <cluster name>
master: true
worker: false

Missing k3s-version & k3s-channel & etc... when execute start & stop & join commands.

autok3s list should add a status field to mark the cluster status

 ✘ ksd@Ksd  sudo autok3s list

     NAME        REGION    PROVIDER  MASTERS  WORKERS
  ksd-autok3s  cn-beijing  alibaba   1        2

update to v1.19.2-k3s1

Is it possible to connect to a node through autok3s ssh node? Just like docker-machine ssh

If the host is a centos7.x operating system, after executing curl https://get.docker.com | VERSION=19.03 sh -s - --mirror Aliyun, the docker server will not start automatically, and the docker service will not be automatically started after the host is started, so when starting for the first time, you need to add the following two commands:

systemctl start docker
systemctl enable docker

Autok3s Startup Success Logs:

=========================== Prompt Info ===========================
Use 'autok3s kubectl config use-context ksd-autok3s.cn-beijing'
Use 'autok3s kubectl get pods -A' get POD status%

After the K3S cluster is successfully started, some Prompt Info will be returned. These prompt messages should be added with the sudo command, for example:

=========================== Prompt Info ===========================
Use 'sudo autok3s kubectl config use-context ksd-autok3s.cn-beijing'
Use 'sudo autok3s kubectl get pods -A' get POD status%

CNI Installer Image zhenyangzhao/cni-installer:v0.8.6 Need Support arm/arm64/amd64

Related:

https://github.com/Jason-ZW/cni-installer

• Unit Test
• Integration Test
• e2e Test

I hope there is a command to view the detailed information of the cluster created by autok3s, such as:

1. What are the nodes and roles
2. Public IP and Intranet IP of the node
3. Node ID
4. docker version
5. Node status

Now autok3s only supports EIP, but it should also support ECS public IP, because some users may use the cluster for a short time, so there is no need to use EIP

It is recommended to add a confirmation operation when operating autok3s delete, otherwise, it is easy to misuse

 ksd@Ksd  ~  sudo autok3s delete \
    --provider alibaba \
    --region cn-beijing \
    --name ksd-autok3s
INFO[0000] [alibaba] executing delete cluster logic...
INFO[0032] [alibaba] successfully excuted delete cluster logic

Using upx compress binaries.

Need to support alibaba cloud-controller-manager component

Related:

github.com/Jason-ZW/autok3s-charts

Compatible

Need both compatible 1.18.x-k3sx & 1.19.x-k3sx

Support Aliyun available zone. e.g. --zone cn-hangzhou-l

Normal users should be supported to run Autok3s

$ autok3s version
ERRO[0000] autok3s: need to be root

Currently, the virtual machine uses the temporarily assigned IP and we need to support the fixed IP.

Need to support the command line：

• autok3s version
• autok3s list
• autok3s create
• autok3s delete
• autok3s stop
• autok3s start
• autok3s remove
• autok3s join
• autok3s ssh
• autok3s kubectl
• autok3s completion

Currently, when choosing the Aliyun Provider, the image pull speed is very slow, and we need to provide a way to solve this problem.

Known Issues

Containerd: Image pull stuck when using a registry endpoint list

Need Version

v1.19.x-k3s1

Description

Run autok3s delete --provider alibaba --name xxx -f. Delete successful but instance still exist in Aliyun console.

This scenario only occurs when creating a cluster with --terway "eni" and associate pod with ENI IP.

Need to support UI components for K3s cluster

• dashboard
• octopus

In the same k3s cluster, when two or more master or worker nodes are created, the Alibaba Cloud instance names are the same

Worker nodes missing 'cloud-provider=external' flag when ccm enable.

Need to Add More Documents:

• Usage Documents
• Developer Documents