provider/v3 unable to read prior state

[utahcon]

Terraform, Provider, Kubernetes and Helm Versions

Terraform version: 1.10.5
Provider version: 3.0.0-pre1
Kubernetes version: 1.32 (eks.2)

Affected Resource(s)

• helm_release

Terraform Configuration Files

resource "helm_release" "karpenter" {
  chart      = "karpenter"
  name       = "karpenter"
  namespace  = "kube-system"
  version    = "1.1.2"
  repository = "oci://public.ecr.aws/karpenter"

  values = [
    templatefile("${path.module}/karpenter/values.yaml", {
      CLUSTER_NAME       = local.cluster.name
      KARPENTER_ROLE_ARN = data.terraform_remote_state.eks.outputs.karpenter.role.arn
    })
  ]

  wait = true

  timeout = 60
}

Debug Output

https://gist.githubusercontent.com/utahcon/2e12436524c7b79e65cc98cd85248aa3/raw/d20e9016cabe95a4cf61f56776d66baf63eb9794/tfp.log

Steps to Reproduce

1. Change from helm provider 2.17.0 to 3.0.0-pre1
2. terraform init --upgrade
3. Updated kubernetes configuration object, and set parameters for helm_release
4. terraform plan

Expected Behavior

Terraform is expected to pull the previous state, parse the files, and then compare the new configuration for changes.

Actual Behavior

Terraform and the Helm provider are unable to parse the existing state using the new provider

╷
│ Warning: Failed to decode resource from state
│ 
│ Error decoding "helm_release.karpenter" from prior state: missing expected {
╵
╷
│ Error: Unable to Read Previously Saved State for UpgradeResourceState
│ 
│   with helm_release.karpenter,
│   on karpenter.tf line 45, in resource "helm_release" "karpenter":
│   45: resource "helm_release" "karpenter" {
│ 
│ There was an error reading the saved resource state using the current resource schema.
│ 
│ If this resource state was last refreshed with Terraform CLI 0.11 and earlier, it must be refreshed or applied with an older provider version first. If you manually modified the resource state, you will need to manually modify it to
│ match the current resource schema. Otherwise, please report this to the provider developer:
│ 
│ AttributeName("metadata"): invalid JSON, expected "{", got "["

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

[hulakv]

Have the same issue. Any updates?

[LifeMoroz]

The same problem with
Provider version: 3.0.0-pre2

[rosmo]

Saw this error too - looks like there are no StateUpgraders in the v3, so I'm not sure if it's meant to preserve backwards compatibility or not.

[lae]

For this issue in particular, the metadata field happens to be a list in v2 and was retyped into an object/map in v3. Attributes for the most part appear to be the same, except v3 seems to have removed the notes attribute.

terraform-provider-helm/helm/resource_helm_release.go

Lines 348 to 389 in 6528832

	"metadata": schema.SingleNestedAttribute{
	Description: "Status of the deployed release.",
	Computed: true,
	Attributes: map[string]schema.Attribute{
	"app_version": schema.StringAttribute{
	Computed: true,
	Description: "The version number of the application being deployed",
	},
	"chart": schema.StringAttribute{
	Computed: true,
	Description: "The name of the chart",
	},
	"first_deployed": schema.Int64Attribute{
	Computed: true,
	Description: "FirstDeployed is an int64 which represents timestamp when the release was first deployed.",
	},
	"last_deployed": schema.Int64Attribute{
	Computed: true,
	Description: "LastDeployed is an int64 which represents timestamp when the release was last deployed.",
	},
	"name": schema.StringAttribute{
	Computed: true,
	Description: "Name is the name of the release",
	},
	"namespace": schema.StringAttribute{
	Computed: true,
	Description: "Namespace is the kubernetes namespace of the release",
	},
	"revision": schema.Int64Attribute{
	Computed: true,
	Description: "Version is an int32 which represents the version of the release",
	},
	"values": schema.StringAttribute{
	Computed: true,
	Description: "Set of extra values. added to the chart. The sensitive data is cloaked. JSON encoded.",
	},
	"version": schema.StringAttribute{
	Computed: true,
	Description: "A SemVer 2 conformant version string of the chart",
	},
	},
	},

Without investigating any further I'm assuming in v2 metadata was always a single item list (looking at our own environment's state file with several helm releases, all I saw were single item lists), so it wasn't necessary to use a list...in that case, adding a StateUpgrader should be okay, right?

[lae]

postrender appears to be another affected attribute that went through the same type change.

Anyway I tried to manually modify these in the state, minimally, but ran into a crash (still as a result of upgrading the helm provider with existing releases so I'm keeping it in this thread) that seems to be happening elsewhere... https://gist.github.com/lae/e08f6df9851a7f92ca05591a53db980c

[jaylonmcshan19-x]

The bug has been resolved with this PR #1633 .
Thank you for creating the issue @utahcon :)

[dima-sh-papaya]

getting the same error with 3.0.2 provider version, after updating my configs according to changelog, and running init with '-upgrade' so it'll use the latest provider (I verified the lock file had 3.0.2 at that moment). are there any additional steps for some state migration or it was supposed to run correctly? not sure where to look for the reason of that error, if it is considered solved in that version of provider, should i pull the state and modify anything manually?

[cgsyam]

I am trying to upgrade provider version from 3.0.0-pre2 to 3.0.2 and facing the same issue.