verizon / 5gedgetutorials Goto Github PK

Goal:
I am trying to launch EKS cluster on wavelength.

Issue Description
After creating the cloudformation stack using the template, the worker nodes aren't joining to the created eks stack. I also verified on the aws dashboard that eks cluster didn't have any nodes attached to it and it seemed to have all required permissions.

Reproducing the error:
I launched the cloudformation template using the cli command provided in the readme and was able to successfuly create the cloudformation stacks.

aws cloudformation create-stack \
--stack-name myWavelengthStack \
--template-body <selected-cfn-file> \
--parameters ParameterKey=EnvironmentName,ParameterValue=WavelengthCloudFormation \
--capabilities CAPABILITY_NAMED_IAM

The small modifications I made to the template was to change

1. Wavelength zone geo to default to us-west-2-wl1-sfo-wlz-1
2. ParentRegion(1/2)Geo default to us-west-2a/b

I then got the kubeconfig context

aws eks update-kubeconfig --name {my_cluster_name} --kubeconfig <local file>

Then using the aws-auth-cm.yml setup in README, I was able to apply the configmap to the cluster.

At this point, I am able to reach the cluster and the aws-auth config map is applied

$ kubectl get configmaps -n kube-system
NAME                                 DATA   AGE
aws-auth                             1      11h
coredns                              1      12h
cp-vpc-resource-controller           0      12h
eks-certificates-controller          0      12h
extension-apiserver-authentication   6      12h
kube-proxy                           1      12h
kube-proxy-config                    1      12h

Querying the events provides failed scheduling due to no nodes

$ kubectl get events --all-namespaces 
NAMESPACE     LAST SEEN   TYPE      REASON             OBJECT                                            MESSAGE
kube-system   2m40s       Warning   FailedScheduling   pod/coredns-5946c5d67c-p9mf8                      no nodes available to schedule pods
kube-system   2m40s       Warning   FailedScheduling   pod/coredns-5946c5d67c-zlrxw                      no nodes available to schedule pods

With no nodes returning:

$ kubectl get nodes
No resources found in default namespace.

As a side note, I am able to launch regular eks instances using eksctl.

The README does mention attaching carrier ips, but I'm not sure how/why this needs to be done and if this is the issue. If this is the issue, is there a way to do it automatically?

For Public Cluster, ECR/EC2 endpoints are Optional but you must manually attach Carrier IP to public-facing instaces

I was also debugging with https://aws.amazon.com/premiumsupport/knowledge-center/eks-worker-nodes-cluster/ and did notice the eks subnets did not have a wavelength zone. I'm not sure if this could be the issue.

Following the instructions in https://github.com/Verizon/5GEdgeTutorials/blob/main/edge-discovery/README.md is successful until the createServiceRegistry() call in Step 2, which returns the following error:

TypeError: createServiceRegistry() missing 1 required positional argument: 'fqdns'.

Any guidance would be appreciated. Thank you.