weaveworks / kubernetes-ami

A simple AMI and CloudFormation for launching Kubernetes on AWS

Get Started with Kubernetes on AWS using CloudFormation

A simple CloudFormation template and AMI builder for running Kubernetes on AWS EC2.

The AMI is built with Packer and includes the Kubernetes packages for installation with kubeadm.

Design

The CloudFormation template creates the following key components:

• Kubernetes master EC2 instance
• Auto Scaling Group for the Kubernetes minions
• Security Group for Kubernetes & pod network comms

Parameters

Required parameters:

• KubeCommunityAMI (AMI identifier)
• KeyName (EC2 SSH key pair identifier)
• NetworkAddon (Can be Weave or NONE, more to be added)

Optional parameters:

• MasterInstanceType (default m4.large)
• NodeInstanceType (default m4.xlarge)
• Nodes (default 3)

Outputs

Once the cluster is running, you need to login to it!

• MasterIP
• LoginToMasterCommand
• GetKubeconfigCommand

Manual Deployment Instructions

There are AMIs published in in all EC2 regions, please consult cloudformation.json for image IDs.

You will need to have an EC2 key in the region where you would like to deploy the cluster.

To create a stack, first clone this repo:

git clone https://github.com/weaveworks/kubernetes-ami
cd kubernetes-ami

Next, copy the command shown above and replace <YOUR_EC2_KEY_NAME> with the name of your SSH key in us-west-2 region.

aws cloudformation create-stack \
    --stack-name KubernetesGettingStarted \
    --region us-west-2 \
    --template-body "file://cloudformation.json" \
    --parameters \
      ParameterKey=KeyName,ParameterValue=<YOUR_EC2_KEY_NAME>

By default a 3-node cluster will be deployed, which takes a few minutes... You can run the following command to check the status of the stack.

> aws --region us-west-2 cloudformation describe-stacks --stack-name KubernetesGettingStarted
{
    "Stacks": [
        {
            "StackId": "arn:aws:cloudformation:us-west-2:992485676579:stack/KubernetesGettingStarted/802a0dad-ad8f-4273-b240-a0f313e1b288",
            "Description": "Getting Started with Kubernetes",
            "Parameters": [
                {
                    "ParameterValue": "<YOUR_EC2_KEY_NAME>",
                    "ParameterKey": "KeyName"
                },
                {
                    "ParameterValue": "m4.large",
                    "ParameterKey": "MasterInstanceType"
                },
                {
                    "ParameterValue": "m4.xlarge",
                    "ParameterKey": "NodeInstanceType"
                },
                {
                    "ParameterValue": "Weave",
                    "ParameterKey": "NetworkAddon"
                },
                {
                    "ParameterValue": "3",
                    "ParameterKey": "Nodes"
                }
            ],
            "Tags": [],
            "Outputs": [
                {
                    "OutputKey": "GetKubeconfigCommand",
                    "OutputValue": "scp -i <YOUR_EC2_KEY_NAME>.pem [email protected]:kubeconfig ./kubeconfig"
                },
                {
                    "OutputKey": "LoginToMasterCommand",
                    "OutputValue": "ssh -i <YOUR_EC2_KEY_NAME>.pem [email protected]"
                },
                {
                    "OutputKey": "MasterIP",
                    "OutputValue": "52.36.245.255"
                }
            ],
            "CreationTime": "2017-02-10T16:14:26.733Z",
            "StackName": "KubernetesGettingStarted",
            "NotificationARNs": [],
            "StackStatus": "CREATE_COMPLETE",
            "DisableRollback": false
        }
    ]
}

As you can see the Outputs section provides a few handy commands you can use to access the cluster.

Have fun using Kubernetes on AWS!

Getting Help

If you have any questions about, feedback for or problems with kubernetes-ami:

• Invite yourself to the Weave Users Slack.
• Ask a question on the #general slack channel.
• File an issue.

Your feedback is always welcome!

kubernetes-ami's Issues

Does not start the kubernetes API

Hi weaveworks!

I just wanted to clarify this repository https://github.com/weaveworks/kubernetes-ami does not do any kubeadm init, correct? It's a bit confusing because the README says "Have fun using Kubernetes on AWS!" as if it was ready to use! I think that after the deploy is done, we have a few machines running with the right packages, but we still need to run some commands to configure things and get some specific kubernetes processes running right?

Ingress

Right user can use node ports as the means of exposing their apps to the outside world, or kubectl port-forward for accessing them from their browser.
But it'd be nice if there existed an ingress controller, e.g. Traefik, run it on every node with ELB in front of it.

Add mapping for pod networks

Add a mapping so users can choose weave, calico, flannel and so on, internally selecting the correct URL for the kubectl apply -f $URL

Add instance types as a List

Publish CloudFormation to S3

So it can be used for "1-click" button.

Publish AMIs to all regions

And add an auto-selection menu to CloudFormation.

Make bootstrap token actually random

There is a Lambda-based solution, or we could cheat by hashing and chopping a local ARN, like the stack ID, which is something like arn:aws:cloudformation:us-west-2:376248598259:stack/kube-7/01c8e3f0-ef25-11e6-899f-500c593b9a36.

it's possible for ec2metadata --public-ipv4 to return wrong address

I've reproduced this once, my API server ended-up advertising wrong address. I wonder if this is the case of a) bug in the metadata API – garbage address being returned; b) a race condition – address being return is valid at the time and is simply the one that we get EIP that we have asked for.