kops export kubecfg server name

[olu-alabi]

Hi,

I just upgraded my kops from version 1.25.4 to 1.25.16. I have noticed that when I run kops export kubecfg --admin, the server name in my kube config is now the long elb name. Is there a flag or a way to return it to the kops 1.25 server name?

kops 1.26
server: https://api-xxxxxxxxx-xxxxxxx-1234567890.eu-west-2.elb.amazonaws.com/

kubectl cluster-info
Kubernetes control plane is running at https://api-xxxxxxxxx-xxxxxxx-1234567890.eu-west-2.elb.amazonaws.com

Kops 1.25

server: https://api/.<cluster_name>

kubectl cluster-info
Kubernetes control plane is running at https://api/.<cluster_name>

[hakman]

Hi @olu-alabi, I think the flag you are looking for is --api-server, but it's available starting kOps 1.33.

[olu-alabi]

Hi @olu-alabi, I think the flag you are looking for is --api-server, but it's available starting kOps 1.33.

Thanks @hakman . So basically no way to force the kops 1.25 behaviour until I am able to upgrade to kops 1.33

[olu-alabi]

What I meant is that I want Kops to build the cluster using the kops 1.25 behaviour(using DNS name of server address) instead of kops 1.26 behaviour (using the address of the load balancer)

[hakman]

Could you explain the reason for the ask?

[olu-alabi]

Thanks. @hakman I have a lot of jobs that uses the api endpoint DNS name. In 1.25 this was api.cluster name but in 1.26 this is now api-xxxxxxxxx-xxxxxxx-1234567890.eu-west-2.elb.amazonaws.com. I will have to do a lot of reconstructions on those jobs to get them working again on kops 1.26. Moreover, the kops 1.26 release note does not mention this at all. so I am also wondering what/where this change was introduced.

[hakman]

The jobs can still use that DNS name, it should not change, Just the kubeconfig points to a valid IP.
This is a change that was introduced to fix issues with misconfigured DNS. It may not be part of the release notes, as it's not something anyone should rely on.

[olu-alabi]

Thank you @hakman . I was actually wondering if there is a simple flag I could pass into my cluster.yaml to force the old 1.25 behaviour? So i guess the only way around it is to set the kops export kubeconfig to point to a valid IP or set the server to the api DNS name.