pulumi: Pulumi v3.10.1+ hangs forever when previewing

pulumi preview in v3.10.1+ hangs forever if you use a pulumi_eks.Cluster as the parent of another resource. Here’s the smallest demonstration I could come up with:

cluster = eks.Cluster(
    "example",
    vpc_id="doesntmatter",
    skip_default_node_group=True,
)

aws.eks.NodeGroup(
    "example",
    cluster_name="example",
    instance_types=["t3.micro"],
    node_role_arn=cluster.instance_roles[0].arn,
    subnet_ids=["doesntmatter"],
    scaling_config=aws.eks.NodeGroupScalingConfigArgs(
        desired_size=1,
        max_size=1,
        min_size=1,
    ),
    opts=pulumi.ResourceOptions(
        # Comment out the following line to make preview work.
        parent=cluster,
    ),
)

I’ve put together a Git repository that packages this reproducer up at https://github.com/benesch/pulumi-bug-eks. Instructions are in the README there.

I’m almost positive that https://github.com/pulumi/pulumi/commit/ecb98b66fd45c1cd5958bfd912eee8b8ce51a09d is the culprit, based on the fact that it is the only commit in v3.10.1 that touches the Python SDK. From a quick skim, I wouldn’t be surprised if there’s additional trickiness that needs to happen when a remote resource is used as the parent of a local resource.

Not setting the cluster as the parent of the node group is a perfectly fine workaround for us for now. Just lost a fair amount of hair tracking this one down. 😄

cc @pgavlin