Helm manifest diffs aren't generated for OCI charts about terraform-provider-helm

[mmckeen]

I also notice in other code paths, we create a cpo object via

terraform-provider-helm/helm/resource_release.go

Line 555 in 8da8621

client := action.NewInstall(actionConfig)

(actions.NewUpgrade/NewInstall) which is responsible for initializing the repository client. That doesn't seem to happen within the resourceDiff codepath and instead the cpo object is initialized empty

terraform-provider-helm/helm/resource_release.go

Line 860 in 8da8621

cpo, chartName, err := chartPathOptions(d, m, &chartPathOpts)

and thus a repository client is never present for the getChart call.

[mmckeen]

I'd also ask that we make

terraform-provider-helm/helm/resource_release.go

Line 868 in 8da8621

return nil

return the error rather than silently failing.

[mmckeen]

Currently working on a draft PR to fix this.

[BBBmau]

Hello @mmckeen, can you provide a terraform config that reproduces the output that you're experiencing? Your provided config is lacking information in order to be reproducible.

[mmckeen]

Hello @mmckeen, can you provide a terraform config that reproduces the output that you're experiencing? Your provided config is lacking information in order to be reproducible.

Hey @BBBmau 👋, the issue will show up whenever the stored manifest for the chart changes as part of the plan, I'm not able to fully produce a config with the issue since we're using private charts but any change to the resource that changes the applied manifest should be able to reproduce the issue.

[sbeginCoveo]

We are having the same issue.

I tested using the same chart, with identical config, on two resources. One is coming from a public github repo, the other is from our private ECR using OCI. We pushed the chart as-is in our private ECR. The only one providing a manifest diff is the public one.

resource "helm_release" "prometheus_operator" {
  name                = "prometheus-operator"
  chart               = "oci://###.dkr.ecr.us-east-1.amazonaws.com/helmchart/kube-prometheus-stack"
  repository_username = var.private_ecr_creds.username
  repository_password = var.private_ecr_creds.password
  namespace           = var.namespace
  version             = "54.2.2"

  values = sensitive([local.prometheus_values])
}

resource "helm_release" "prometheus_operator" {
  name       = "prometheus-operator"
  repository = "https://prometheus-community.github.io/helm-charts"
  chart      = "kube-prometheus-stack"
  namespace  = var.namespace
  version    = "54.2.2"

  values = sensitive([local.prometheus_values])
}

In these examples, local.prometheus_values is a templated yaml values file using templatefile. Its set sensitive just because we don't want to see the huge values diff as it has low value for us. Manifests, though are high value in the diff

Hope this helps with reproduction