[HELP] Configure AWS Cloud Controller Manager

[codersergg]

I successfully deploy the cluster using the UI interface. I also successfully deploy the cluster using the generated CLI command. However, when I add the following settings to the command:

autok3s -d create -p aws \
    ... \
    --cloud-controller-manager \
    --iam-instance-profile-control my-iam-policy-for-control-plane \
    --iam-instance-profile-worker my-iam-policy-for-node

after configuring the policies, I receive an error: "Job for k3s-agent.service failed because the control process exited with error code."

Setting the flag --disable=servicelb does not resolve the issue.

[JacieChao]

Hi @codersergg Thanks for your feedback. Could you please help to provide more information about

• The K3s cluster version you deployed
• Will the cluster deploy successfully with cloud-controller-manager enabled with only one master node?

Also, this workaround may help.

1. Create a K3s cluster with only 1 master node and set the rollback to false to keep the instance running even AutoK3s failed to deploy the K3s cluster for debugging reasons.

   autok3s -d create -p aws \
   ... \
   --master 1 \
   --cloud-controller-manager \
   --iam-instance-profile-control my-iam-policy-for-control-plane \
   --iam-instance-profile-worker my-iam-policy-for-node \
   --rollback false

2. Check the aws-cloud-controller-manager pod from the AWS instance to check if there are logs like

   routecontroller.go:52] Couldn't reconcile node routes: error listing routes: unable to find route table for AWS cluster: kubernetes

3. Using the kubectl edit to add arg --configure-cloud-routes=false for the aws-cloud-controller-manager Daemonset
4. Check the status of the K3s cluster.

[codersergg]

1. I used the latest version.
2. It worked with a single node.

Following the instructions, I disabled configure-cloud-routes, and after that, I was able to successfully join the worker node. This looks a bit odd. Why wasn't cloud routes configured automatically?

[JacieChao]

I'm glad that the workaround helps and I can confirm this is a bug after upgrading aws-cloud-provider to v1.27.1.

To use cloud routes, we need to tag the VPC route table with the tag kubernetes.io/cluster/<cluster-id>

This looks a bit odd. Why wasn't cloud routes configured automatically?

Yes. The cloud routes need to be configured automatically. We will fix this in the next version.