Azure Container Service (AKS) kubeconfig file outdated - azure

Question

I am learning about K8s and did setup a release pipeline with a kubectl apply. I've setup the AKS cluster via Terraform and on the first run all seemed fine. Once I destroyed the cluster I reran the pipeline, I get issues which I believe are related to the kubeconfig file mentioned in the exception. I tried the cloud shell etc. to get to the file or reset it but I wasn't succesful. How can I get back to a clean state?
2020-12-09T09:08:51.7047177Z ##[section]Starting: kubectl apply
2020-12-09T09:08:51.7482440Z ==============================================================================
2020-12-09T09:08:51.7483217Z Task : Kubectl
2020-12-09T09:08:51.7483729Z Description : Deploy, configure, update a Kubernetes cluster in Azure Container Service by running kubectl commands
2020-12-09T09:08:51.7484058Z Version : 0.177.0
2020-12-09T09:08:51.7484996Z Author : Microsoft Corporation
2020-12-09T09:08:51.7485587Z Help : https://learn.microsoft.com/azure/devops/pipelines/tasks/deploy/kubernetes
2020-12-09T09:08:51.7485955Z ==============================================================================
2020-12-09T09:08:52.7640528Z [command]C:\ProgramData\Chocolatey\bin\kubectl.exe --kubeconfig D:\a\_temp\kubectlTask\1607504932712\config apply -f D:\a\r1\a/medquality-cordapp/k8s
2020-12-09T09:08:54.1555570Z Unable to connect to the server: dial tcp: lookup mq-k8s-dfee38f6.hcp.switzerlandnorth.azmk8s.io: no such host
2020-12-09T09:08:54.1798118Z ##[error]The process 'C:\ProgramData\Chocolatey\bin\kubectl.exe' failed with exit code 1
2020-12-09T09:08:54.1853710Z ##[section]Finishing: kubectl apply
Update, workflow tasks of the release pipeline:
Initially I get the artifact, clone of the repo containing the k8s yamls, then the stage does a kubectl apply.
"workflowTasks": [
{
"environment": {},
"taskId": "cbc316a2-586f-4def-be79-488a1f503564",
"version": "0.*",
"name": "kubectl apply",
"refName": "",
"enabled": true,
"alwaysRun": false,
"continueOnError": false,
"timeoutInMinutes": 0,
"definitionType": null,
"overrideInputs": {},
"condition": "succeeded()",
"inputs": {
"kubernetesServiceEndpoint": "82e5971b-9ac6-42c6-ac43-211d2f6b60e4",
"namespace": "",
"command": "apply",
"useConfigurationFile": "false",
"configuration": "",
"arguments": "-f $(System.DefaultWorkingDirectory)/medquality-cordapp/k8s",
"secretType": "dockerRegistry",
"secretArguments": "",
"containerRegistryType": "Azure Container Registry",
"dockerRegistryEndpoint": "",
"azureSubscriptionEndpoint": "",
"azureContainerRegistry": "",
"secretName": "",
"forceUpdate": "true",
"configMapName": "",
"forceUpdateConfigMap": "false",
"useConfigMapFile": "false",
"configMapFile": "",
"configMapArguments": "",
"versionOrLocation": "version",
"versionSpec": "1.7.0",
"checkLatest": "false",
"specifyLocation": "",
"cwd": "$(System.DefaultWorkingDirectory)",
"outputFormat": "json",
"kubectlOutput": ""
}
}
]
```

Answer 1

I can see you are using kubernetesServiceEndpoint as the Service connection type in Kubectl task.
Once I destroyed the cluster I reran the pipeline, I get issues....
If the cluster was destroyed. The kubernetesServiceEndpoint in azure devops is still connected to the origin cluster. Kubectl task which using the origin kubernetesServiceEndpoint is still looking for the old cluster. And it will fail with above error, since the old cluster was destroyed.
You can fix this issue by updating the kubernetesServiceEndpoint in azure devops with the newly created cluster:
Go to Azure devops Project settings-->Service connections--> Find your Kubernetes Service connection-->Click Edit to update the configuration.
But if your kubernete cluster gets destroyed and recreated frequently. I would suggest using Azure Resource Manager as the Service connection type to connect to the cluster in Kubectl task. See below screenshot.
By using azureSubscriptionEndpoint and specifying azureResourceGroup, if only the cluster's name doesnot change, It doesnot matter how many times the cluster is recreated.
See document to create an Azure Resource Manager service connection

Answer 2

When you destroy and reprovision AKS cluster the kube API URL and some other things change, but as you found out, nothing updates this automatically on your configured clients.
What I do to get access new and reprovisioned AKS clusters is :
az aks get-credentials --subscription <sub> -g <rg> -n <aksname> -a --overwrite