-
Notifications
You must be signed in to change notification settings - Fork 300
Description
I have a use case to deploy a custom kubernetes cluster on VMs in azure local (previously azure stack hci).
I am able to deploy the vms and bootstrap kubernetes, but I am having trouble getting "cloud-provider-azure" to work properly.
There are documentation here that describe how to setup configuration for "Azure Stack". Is "Azure Stack" in this instance the same as Azure Local or Azure Stack HCI?
I have been able to follow those documentation and get the "cloud-controller-manager" and "cloud-node-manager" running on my cluster, but I did have to make some manual changes specifically to the helm chart.
The docs indicate that you must provide an "environment file" that defines the endpoints for the environment to the cloud controller manager, but there is no way to do that in the helm chart right now. I would expect the ability to add extra volumes to the helm chart in order to mount a secret containing that environment to the cloud-controller-manager pod and then also be able to define extra environment variables to set the "AZURE_ENVIRONMENT_FILE" as required by the documentation. But that does not seem to be available.
Moreover, once I got the cloud-controller-manager and cloud-node-manager running after the changes to the deployment manifests, the cloud-controller-manager did not remove the "node.cloudprovider.kubernetes.io/uninitialized" taint thus preventing certain pods from being scheduled.
Is cloud-provider-azure a verified implementation of CPI for a Bring Your Own VM cluster on Azure Local? If so, can the helm chart be updated to support the requirements found in the documentation?