This might not be a bug, but a bit more information is needed for using ubuntu-eks in AWS EKS cluster

[Jianhua Wu]

AWS EKS Cluster: 1.15
Ubuntu EKS Image: ubuntu-eks/k8s_1.15/images/hvm-ssd/ubuntu-bionic-18.04-amd64-server-20200318.1

Issue Symptom:
My automation code for provisioning EKS work node groups with AWS default image works;

My automation code for provisioning EKS work node groups with Ubuntu EKS image (ami-02d443718ad8d42d8) does NOT work;

The command "kubectl get nodes" does not return anything if I provisioned my work node group with Ubuntu-EKS image.

From EC2 web console, there is only one private IP address associated to my Ubuntu worker node EC2 server; however, there are multiple private IP address (and even multiple secondary IP address) associated with my AWS default node image. (not sure if it was designed to work this way)

Sos report is attached in this ticket.

During the launch process of provisioning EC2 server with Ubuntu-EKS image, is there any external endpoint(S3 bucket, docker registry, and alike) that Ubuntu-EKS EC2 server tries to contact in order to download any package which is required for register Ubuntu EKS node with EKS cluster 1.15?

Thanks

[Jianhua Wu (wujianhua)]

systemctl status kubelet

The output from this command returned nothing. and I could not find a kubelet.service file under the path /etc/systemd/system

This might explain why my Ubuntu-EKS worker nodes are not registered in my EKS cluster 1.15.

I checked other requirements from the link below, and it seems the only missing component on my Ubuntu-EKS EC2 server is kubelet service file.
https://medium.com/getamis/customize-aws-eks-worker-node-ami-d7b3d203051

[Cody Shepherd (codyshepherd)]

Hi there. This looks to be a duplicate of LP: #1869562. Can you re-confirm with a 1.15 image with serial 20200406.1 or later, as listed on the Ubuntu EKS webpage: https://cloud-images.ubuntu.com/docs/aws/eks/ ?

Thank you.

[Cody Shepherd (codyshepherd)]

Also, the cluster and nodegroup workflow requirements for using Ubuntu with EKS may not be obvious due to recent changes to AWS docs.

We recommend following one of two strategies:
1. Using the eksctl tool [1] to stand up the cluster and nodegroup using a command such as:

eksctl create cluster \
  --name <cluster name> \
  --version 1.15 \
  --nodegroup-name test-workers \
  --node-type <instance shape> \
  --nodes <n> \
  --nodes-min <n> \
  --nodes-max <n> \
  --node-ami <ami-id> \
  --node-ami-family Ubuntu1804 \
  --region <region> \
  --ssh-access \
  --ssh-public-key <key registered w/ AWS>

2. Launch an EKS cluster and self-managed nodegroup manually:
  a. Launch a cluster as described in the AWS docs [2]. This should include the step to run `aws eks update-kubeconfig`
  b. Launch a self-managed nodegroup following the instructions for launching self-managed nodes in the AWS docs [3], but substitute the desired Ubuntu ami-id for the NodeImageId field.
  c. Follow the step in the above docs to download, edit, and apply 'aws-auth-cm.yaml.'

[1] https://eksctl.io/
[2] https://docs.aws.amazon.com/eks/latest/userguide/getting-started-console.html
[3] https://docs.aws.amazon.com/eks/latest/userguide/launch-workers.html#self-managed-nodes

[Jianhua Wu (wujianhua)]

Hi Cody
Thank you for sharing the info.

With the new image, my Ubuntu-EKS worker node is now registered with my EKS 1.15.

Best