Hi All, I have created an EKS Cluster and exporting the cluster object something like this in one of the stacks

self.eks_cluster     = eks.Cluster()
pulumi.export('eks_cluster', self.eks_cluster)

From a different stack I am trying to get the cluster object to create ManagedNodeGroups

eks_cluster: eks.Cluster = infrastructure.get_output("eks_cluster")

It gives me this error. It only returns a dict object.

AttributeError: 'dict' object has no attribute 'urn'

Is is possible to get the exported values as Objects and can be casted? any thoughts?

I had the same problem a while ago. As far as I know, it's not possible to export and import the entire

eks.Cluster()

component resource. You can only export values. There is also no equivalent to

aws.eks.getCluster()

(docs) for the pulumi_eks module, see this issue. What you can do, however, is using

aws.eks.NodeGroup

(docs) to add node groups to your cluster. This only requires the cluster's name and will also create what EKS calls a managed node group. You can see in the source code that

eks.ManagedNodeGroup

wraps

aws.eks.NodeGroup

and doesn't do a whole lot more, aside from reading some details from the

eks.Cluster

resource rather than you having to pass them explicitly.

I was thinking of adding NodeGroups, but just wanted to leave the AMI for AWS to handle. I think this is the only way to workaround this problem. Let me try NodeGroup. Thank you

You don't have to specify an AMI when creating an

aws.eks.NodeGroup

, both

ami_type

and

release_version

are optional parameters. This is exactly the same as for

eks.ManagedNodeGroup

, which simply passes these parameters down to the underlying

aws.eks.NodeGroup

resource (here). The naming is unfortunately really misleading, you'll get the exact same kind of managed node group in both cases.

Seems like. wow, it worked!! Thank you!!

Glad I could help!

Hi @modern-zebra-45309, I need another help. I am exporting the kubeconfig from one stack something like this,

pulumi.export('kubeconfig', self.eks_cluster.kubeconfig)

and then while deploying the helm chart on a different stack using the Stack reference to get this value and add a new provider:

StackReference(                                stack_name = f"organization/core/{get_stack()}"
                                )
eks_kubeconfig            = _sync_await(infrastructure.get_output_details("kubeconfig")).value

eks_cluster_provider    = Provider(
                cluster     = f"{get_stack()}-{subsystem_name}-core-{name}-cluster",
                kubeconfig  = eks_kubeconfig
        )

In the helm chart Release I am passing the following option:

opts            = pulumi.ResourceOptions(
                                               
                                                provider    = eks_cluster_provider
                                        )

You don't show your imports and only some of the variable names, so it's a bit hard to tell. You do not have to specify the cluster when instantiating the provider. Handing over a kubeconfig is sufficient:

import pulumi
import pulumi_eks as eks
import pulumi_kubernetes as k8s

# first program

cluster = eks.Cluster(...)
pulumi.export("kubeconfig", cluster.kubeconfig)

# second program

first_stack = pulumi.StackReference(...)
k8s_provider = k8s.Provider("k8s-provider", kubeconfig=first_stack.get_output("kubeconfig"))

Hope that helps!