currently im getting around it by calling kubernetes api directly to remove the existing csi driver like :-

configuration = kubernetes.config.load_kube_config(config_file='~/.kube/config', context='arn:aws:eks:us-east-1:6299999991:cluster/itplat-eks-cluster')
    with kubernetes.client.ApiClient(configuration) as api_client:
        k8s_api_instance = kubernetes.client.StorageV1beta1Api(api_client)
        csi_read_response = k8s_api_instance.read_csi_driver("<http://efs.csi.aws.com|efs.csi.aws.com>")
        print(f'the efs.csi api_version was {csi_read_response.api_version}')
        if csi_read_response.api_version == '<http://storage.k8s.io/v1|storage.k8s.io/v1>':
            csi_delete_response = k8s_api_instance.delete_csi_driver("<http://efs.csi.aws.com|efs.csi.aws.com>")

but it feels like a super shitty workaround because pulumi isnt supporting helm hooks

so back on this… i would like to give some context. we are creating a single pulumi stack which deploys all the things needed for an eks cluster, then creates the cluster, then populates it with essential components like crossplane pieces and CSI drivers and etc that our application_service_helm charts require to pre-exist

<https://github.com/kubernetes-sigs/aws-efs-csi-driver>

helm deployment uses a preinstall hook to remove the existing efs csi driver and install it’s own which does not work in pulumi because https://github.com/pulumi/pulumi-kubernetes/issues/555

so im trying to work around/emulate that csi driver removal using

csi_delete_response = k8s_api_instance.delete_csi_driver("<http://efs.csi.aws.com|efs.csi.aws.com>")

but im having issue referencing the kube/config created in the same stack

trying things like

configuration = kubernetes.config.load_kube_config(config_file='~/.kube/config.pulumi', context='arn:aws:eks:us-east-1:6299999521:cluster/itplat-eks-cluster')

@billowy-army-68599 im using the kubeconfig code at

<https://github.com/pulumi/examples/blob/master/aws-py-eks/utils.py>

which you pointed me at… any suggestions about how to pass that same config resource to

kubernetes.config.load_kube_config

?

or a suggestion about a better design

the issue im having is that the kube config doesnt exist at preview stage so i cant use it as valid config for kubernetes.config.load_kube_config