Hi, Pulumi Team. Wonderful tool, had fun exploring it’s abilities. Struggled with the helm chart installation, specifically kube-prometheus-stack. Following code fails to create entities based on CRD’s, I guess chart doesn’t install them.

from pulumi_kubernetes.helm.v3 import Chart, ChartOpts, FetchOpts

# ...

def remove_status(obj, opts):
    if obj["kind"] == "CustomResourceDefinition":
        del obj["status"]

prometheus_stack_chart = Chart(
    "prometheus-stack",
    ChartOpts(
        chart="kube-prometheus-stack",
        transformations=[remove_status],
        version="34.8.0",
        fetch_opts=FetchOpts(
            repo="<https://prometheus-community.github.io/helm-charts>",
        ),
        namespace=monitoring_namespace.id,
        values=values,
        skip_await=False
    ),
    pulumi.ResourceOptions(depends_on=[monitoring_namespace, cluster])
)

I may notice the

remove_status

hack (I’ve found on the Internet), without it installation fails before the beginning, ’cause somewhere inside it tries to pass that

status

argument to a function that doesn’t expect it. Am I doing something wrong, or am I just unlucky to need the Chart that is trouble to install with pulumi? Do you have any ideas or suggestions?

Can you confirm that CRD are not installed once the following code is executed?

Yes, they were partially installed. I saw only one out of 8.

I see. I’ll try to repro it on my side but I’m pretty swamped with work today.

I had some problems with a chart that required helm hooks and the solution was to use

pulumi_kubernetes.helm.v3.Release

instead of

Chart

@great-sunset-355 might be on to something. I remember now that pulumi provided (later) a dedicated support for helm releases. That differs a lot compared to the Chart resource that basically renders the helm chart and applies the generated manifests one by one (without creating a helm release).

hm, thanks a lot, I’ll try to use

Release

@victorious-continent-984 any success with the

HelmRelease

approach? I'm running into the following problem:

Use Helm CLI to investigate.: failed to become available within allocated timeout. Error: Helm Release monitoring/kube-prometheus: unable to build kubernetes objects from new release manifest: unable to recognize "":

my code:

new k8s.helm.v3.Release(
      `${name}-kube-prometheus`,
      {
        chart: 'kube-prometheus-stack',
        version: '37.3.0',
        namespace: this.namespace,
        createNamespace: false,
        skipCrds: true,
        repositoryOpts: { repo: '<https://prometheus-community.github.io/helm-charts>' },
      },
      { ...this.OPTS, dependsOn: namespace },
    );

I was trying to play with the

skipAwait

flag but it doesn't change the result. CRDs are all installed correctly. Also setting a

name

for the

HelmRelease

didn't change anything. Any ideas?

Release resource worked fine for me. Some times it’s better to remove the helm release with CLI and create it again with pulumi.

I will try again with a clean state - regarding the

dependsOn

key: documentation allows both array and single item values

/**
     * An optional additional explicit dependencies on other resources.
     */
    dependsOn?: Input<Input<Resource>[]> | Input<Resource>;

the k8s provider is set within the

this.OPTS

object and also works fine for any other kubernetes resource I'm spinning up using pulumi