`duplicate entry for /v1, Kind=APIResourceList` error in v3.1.0 when running `terraform plan` for an existing helm_release with CRDs

[sebbes]

Terraform, Provider, Kubernetes and Helm Versions

Terraform v1.13.4
Provider: hashicorp/helm v3.1.0
Kubernetes Version:1.33.5
Helm Version: "v3.19.

Affected Resource(s)

helm_release

Terraform Configuration Files

resource "helm_release" "cert_manager" {
  name       = "cert-manager"
  repository = "https://charts.jetstack.io"
  chart      = "cert-manager"
  version    = "v1.19.1"
  namespace  = "cert-manager"
  skip_crds  = false

  atomic          = true
  cleanup_on_fail = true
  max_history     = 10

  values = [
    <<EOF
crds:
  enabled: true
serviceAccount:
  annotations:
    iam.gke.io/gcp-service-account: "cert-manager@my-project.iam.gserviceaccount.com"
featureGates: "ServerSideApply=true"
EOF
  ]
}

Debug Output

---

Panic Output

---

Steps to Reproduce

1. terraform plan

Expected Behavior

Should be able to update an existing Helm release

Actual Behavior

 Error: Parser Error
 
   with helm_release.cert_manager,
   on cert_manager.tf line 1, in resource "helm_release" "cert_manager":
    1: resource "helm_release" "cert_manager" {
 
 duplicate entry for /v1, Kind=APIResourceList


 Error: Error setting release attributes
 
   with helm_release.cert_manager,
   on cert_manager.tf line 1, in resource "helm_release" "cert_manager":
    1: resource "helm_release" "cert_manager" {
 
 Unable to set attributes for helm release cert-manager

Important Factoids

K8s cluster runs in GKE.

When running the same code with provider version 3.0.2 it works just fine. In fact, I have the same issue with nginx-ingress, where I also install CRDs. Resources without CRD installations are not affected.

References

---

Community Note

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment

[devopsinfoltd]

We have started looking the same error since yesterday, please push the fix

[jrhouston]

Are you using the manifest experiment:

experiments {
    manifest = true
}

In your provider block?

It looks like this change: https://github.com/hashicorp/terraform-provider-helm/pull/1693/files#diff-c548ca21c915d6a4918945eb9741a292966d795ca0a4e9e6dfa4f8c8106f5691R1677-R1692

Introduced some code that makes some calls to the API to populate the a list of resources, and this is triggering a nasty bug where NewGVKParser errors if there's duplicate entries in the OpenAPI spec.

Related: kubernetes/kubernetes#128201
Related: kubernetes/kubernetes#103597

[sebbes]

@jrhouston Yes, we are using

experiments {
    manifest = true
}

[jriguera]

We are facing exactly this issue, with same versions and manifest = true also running in GKE

[zhimsel]

Any updates on this investigation?

[m00lecule]

Facing similar issues on EKS 1.33+

[KamilPacanek]

Bump, same error.

• K8s 1.33.5
• Terraform 1.14.3
• TF Helm provider 3.1.1

Downgrading helm provider to 3.0.2 resolves issue.

[eklesel]

Would love to see this picked up, as #1713 is a much feature for us but we can't upgrade because of this error. We also use the manifests experiment which is critical to our review process.

[zhimsel]

This seems to have been fixed in 3.1.2, but I'm also seeing spurrious "known after apply" on all fields in a helm_release resource. Not sure if that's related or not.