I'm struggling with create and destroy order for my k8s stack. I'm trying to build a stack that launches eks, then installs the aws LB controller using helm (composite resource), and then create an ingress, which uses the controller to create an ALB. The problem is that I cannot get the ingress creation to wait until the Deployment deployed by the Helm chart is created and initialized, and so the ingress creation often fails. And I cannot get the Deployment teardown to wait until the Ingress is torn down, so the Ingress delete usually blocks indefinitely. I have code that looks like this:

lb_controller_chart = k8s.helm.v3.Chart(
    "albctl",
    k8s.helm.v3.ChartOpts(
        chart="aws-load-balancer-controller",
        fetch_opts=k8s.helm.v3.FetchOpts(
            repo="<https://aws.github.io/eks-charts>"
        ),
        namespace=alb_controller_sa_namespace,
        values={
            "region": "us-west-2",
            "serviceAccount": {
                "name": alb_controller_sa,
                "create": False,
            },
            "vpcId": eks_cluster.eks_cluster.vpc_config.vpc_id.apply(lambda c: c),
            "clusterName": eks_cluster.eks_cluster.name.apply(lambda c: c),
        },
    ),
    opts=pulumi.ResourceOptions(provider=eks_provider)
)
...
ingress = k8s.networking.v1.Ingress(
    "myingress",
    metadata=k8s.meta.v1.ObjectMetaArgs(
        name="myingress",
        annotations={
            "<http://alb.ingress.kubernetes.io/target-type|alb.ingress.kubernetes.io/target-type>": "ip",
            "<http://alb.ingress.kubernetes.io/listen-ports|alb.ingress.kubernetes.io/listen-ports>": "[{\"HTTP\": 80}]",
            "<http://alb.ingress.kubernetes.io/scheme|alb.ingress.kubernetes.io/scheme>": "internet-facing",
            "<http://alb.ingress.kubernetes.io/subnets|alb.ingress.kubernetes.io/subnets>": eks_vpc.public_subnet_ids.apply(lambda ids: ", ".join(ids)),
            "<http://alb.ingress.kubernetes.io/backend-protocol|alb.ingress.kubernetes.io/backend-protocol>": "HTTP"
        }
    ),
    spec=k8s.networking.v1.IngressSpecArgs(
        ingress_class_name="alb",
        rules=[k8s.networking.v1.IngressRuleArgs(
            host="*.helloworld.responsive.dev",
            http=k8s.networking.v1.HTTPIngressRuleValueArgs(
                paths=[k8s.networking.v1.HTTPIngressPathArgs(
                    backend=k8s.networking.v1.IngressBackendArgs(
                        service=k8s.networking.v1.IngressServiceBackendArgs(
                            name="helloworld",
                            port=k8s.networking.v1.ServiceBackendPortArgs(number=8081)
                        )
                    ),
                    path="/",
                    path_type="Prefix"
                )]
            )
        )]
    ),
    opts=pulumi.ResourceOptions(
        provider=eks_provider, depends_on=[lb_controller_chart])
)

At create time, I need the ingress to only get created after the deployment resource that is part of the helm chart is created. However, often times it does not wait and the ingress creation fails. Similarly, at delete time the deployment in the helm chart is deleted before the ingress, and the ingress deletion hangs. I thought setting

depends_on

would cover me here, but it doesn't seem to be working. My pulumi version is

v3.55.0

pulumi_kubernetes is at

pulumi_kubernetes-3.24.1

can you try

depends_on=[b_controller_chart.ready]

? https://www.pulumi.com/registry/packages/kubernetes/api-docs/helm/v3/chart/#depend-on-a-chart-resource

trying it now. Interestingly I don't see that example in the python docs

hmm looks like a bug. You can submit a GitHub Issue but in the meantime does this work for you? It was the workaround in the past:

depends_on=b_controller_chart.resources.apply(lambda resources: list(resources.values())),

https://github.com/pulumi/pulumi-kubernetes/issues/861#issuecomment-901862700

ooh I'll try that next. Right now I'm trying:

depends_on=[lb_controller_chart.get_resource(
            "apps/v1/Deployment", "kube-system/albctl-aws-load-balancer-controller")]

Nice one - TIL^ - I haven't had to do a Chart depending on a Chart but I imagine the workaround is because Pulumi isn't natively iterating thru all the resources, just ensuring the original Chart resource is created/ready in the Stack State and doesn't check all of the children?

yup pretty much