Aws-cli: aws eks update-kubeconfig fails with error TypeError: 'NoneType' object is not iterable

[TechnicalMercenary]

I just received a new MacBook with MacOS Catalina. I used the latest AWS CLI Bundle to install the AWS CLI on the MBP.

`gregory.bonk@greg-bonk-mbp:~/workspace$ unzip awscli-bundle.zip
Archive: awscli-bundle.zip
inflating: awscli-bundle/install
inflating: awscli-bundle/packages/PyYAML-5.2.tar.gz
inflating: awscli-bundle/packages/pyasn1-0.4.8.tar.gz
inflating: awscli-bundle/packages/docutils-0.15.2.tar.gz
inflating: awscli-bundle/packages/botocore-1.14.3.tar.gz
inflating: awscli-bundle/packages/s3transfer-0.3.0.tar.gz
inflating: awscli-bundle/packages/urllib3-1.25.7.tar.gz
inflating: awscli-bundle/packages/python-dateutil-2.8.0.tar.gz
inflating: awscli-bundle/packages/virtualenv-16.7.8.tar.gz
inflating: awscli-bundle/packages/colorama-0.4.1.tar.gz
inflating: awscli-bundle/packages/jmespath-0.9.4.tar.gz
inflating: awscli-bundle/packages/futures-3.3.0.tar.gz
inflating: awscli-bundle/packages/awscli-1.17.3.tar.gz
inflating: awscli-bundle/packages/rsa-3.4.2.tar.gz
inflating: awscli-bundle/packages/six-1.14.0.tar.gz
inflating: awscli-bundle/packages/setup/setuptools_scm-3.3.3.tar.gz
inflating: awscli-bundle/packages/setup/wheel-0.33.6.tar.gz

gregory.bonk@greg-bonk-mbp:~/workspace$ sudo ./awscli-bundle/install -i /usr/local/aws -b /usr/local/bin/aws
Running cmd: /usr/local/opt/python@2/bin/python2.7 virtualenv.py --no-download --python /usr/local/opt/python@2/bin/python2.7 /usr/local/aws
Running cmd: /usr/local/aws/bin/pip install --no-binary :all: --no-cache-dir --no-index --find-links file://. setuptools_scm-3.3.3.tar.gz
Running cmd: /usr/local/aws/bin/pip install --no-binary :all: --no-cache-dir --no-index --find-links file://. wheel-0.33.6.tar.gz
Running cmd: /usr/local/aws/bin/pip install --no-binary :all: --no-build-isolation --no-cache-dir --no-index --find-links file:///Users/gregory.bonk/workspace/devops-application/infrastructure/awscli-bundle/packages awscli-1.17.3.tar.gz`

Checking the install I see this...

gregory.bonk@greg-bonk-mbp:~/workspace$ which aws /usr/local/bin/aws gregory.bonk@greg-bonk-mbp:~/workspace/$ /usr/local/bin/aws --version aws-cli/1.17.3 Python/2.7.17 Darwin/19.2.0 botocore/1.14.3

Then when running a command to get an EKS cluster configuration...

aws eks update-kubeconfig --name my-eks --region us-west-2

This is the error...

'NoneType' object is not iterable

In the debug I see that there is a response, coming back. Appears to be something with parsing it ?

https://gist.github.com/TechnicalMercenary/48f15680f1fcd506b256e84157f08e30

[TechnicalMercenary]

After deleting my existing _~/.kube/config_ file I ran the update-kubeconfig again and now it works perfectly fine.

Perhaps it was an issue with my config's format or something ? Here's what it used to look like...

$ cat ~/.kube/config
apiVersion: v1
clusters: null
contexts: null
current-context: arn:aws:eks:us-west-2:3456345634:cluster/my-eks
kind: Config
preferences: {}
users: null

[gbonk]

I've stumbled upon this again, and I'm going to speculate that the AWS CLI needs to be able to handle config files a little more robustly.

Here's a couple of flows that I've come up with that causes this to fail....

Flow 1

// Start with fresh file
rm ~/.kube/config

// Update with the context you want
aws eks update-kubeconfig --name my-cluster --region us-west-2

// Use KubeCtl to delete the context
kubectl config delete-context arn:aws:eks:us-west-2:000000000000:cluster/my-cluster

// RE-Apply the Config
aws eks update-kubeconfig --name my-cluster --region us-west-2

Tried to insert into contexts, which is a not a

Here is my context file before the Re-Apply

```apiVersion: v1
clusters:

• cluster:
  certificate-authority-data: {...snip...}.sk1.us-west-2.eks.amazonaws.com
  name: arn:aws:eks:us-west-2:000000000000:cluster/my-cluster
  contexts: null
  current-context: arn:aws:eks:us-west-2:000000000000:cluster/my-cluster
  kind: Config
  preferences: {}
  users:
• name: arn:aws:eks:us-west-2:000000000000:cluster/my-cluster
  user:
  exec:
  apiVersion: client.authentication.k8s.io/v1alpha1
  args:
  
  
  
  • --region
  
  
  • us-west-2
  
  
  • eks
  
  
  • get-token
  
  
  • --cluster-name
  
  
  • my-cluster
    
    command: aws
    
    env: null
  
  

# Flow 2

// Start with fresh file
rm ~/.kube/config

// Update with the context you want
aws eks update-kubeconfig --name my-cluster --region us-west-2

// Use KubeCtl to delete the context
kubectl config delete-cluster arn:aws:eks:us-west-2:000000000000:cluster/my-cluster

// RE-Apply the Config
aws eks update-kubeconfig --name my-cluster --region us-west-2

'NoneType' object is not iterable

### Config before re-apply

apiVersion: v1
clusters: null
contexts:

• context:
  cluster: arn:aws:eks:us-west-2:000000000000:cluster/my-cluster
  user: arn:aws:eks:us-west-2:000000000000:cluster/my-cluster
  name: arn:aws:eks:us-west-2:000000000000:cluster/my-cluster
  current-context: arn:aws:eks:us-west-2:000000000000:cluster/my-cluster
  kind: Config
  preferences: {}
  users:
• name: arn:aws:eks:us-west-2:000000000000:cluster/my-cluster
  user:
  exec:
  apiVersion: client.authentication.k8s.io/v1alpha1
  args:
  
  
  
  • --region
  
  
  • us-west-2
  
  
  • eks
  
  
  • get-token
  
  
  • --cluster-name
  
  
  • my-cluster
    
    command: aws
    
    env: null
    
    ```
  
  

[Raffo]

Apparently, the AWS CLI will need to take into account the changed behavior, see how the upstream kubernetes issue was closed: https://github.com/kubernetes/kubernetes/issues/88524

[Jean-Baptiste-Lasselle]

thanks all, i had same issue, solved it with rm ~/.kube/config as well

[kdaily]

Thanks for the report and the detailed descriptions on how to get it to fail. I'll confer with the EKS team on how to handle this, as it is a CLI customization.