unset ovh_cloud_project_kube.version forces recreation

[0xErnie]

Terraform Version

Terraform v1.0.0
on darwin_amd64
+ provider registry.terraform.io/ovh/ovh v0.13.0
+ provider registry.terraform.io/terraform-provider-openstack/openstack v1.42.0

Affected Resource(s)

Please list the resources as a list, for example:

• ovh_cloud_project_kube
• ovh_cloud_project_kube_nodepool

Terraform Configuration Files

provider "ovh" {
  endpoint           = "ovh-eu"
  application_key    = REMOVED
  application_secret = REMOVED
  consumer_key       = REMOVED
}
resource "ovh_cloud_project_kube" "mykube" {
  service_name       = REMOVED
  name               = "management"
  region             = "GRA7"
  private_network_id = REMOVED
}

resource "ovh_cloud_project_kube_nodepool" "pool" {
  service_name   = REMOVED
  kube_id        = ovh_cloud_project_kube.mykube.id
  name           = "pool001"
  flavor_name    = "b2-7"
  anti_affinity  = true
  monthly_billed = false
  desired_nodes  = 3
  max_nodes      = 3
  min_nodes      = 3
}

Expected Behavior

After the first terraform apply there should be no changes to the infrastructure if we run it again.

Actual Behavior

terraform detects that the version of the cluster is "1.20" instead of null so it wants to receate the hole cluster.

  # ovh_cloud_project_kube.mykube must be replaced
-/+ resource "ovh_cloud_project_kube" "mykube" {
      ~ control_plane_is_up_to_date = true -> (known after apply)
      ~ id                          = "REMOVED" -> (known after apply)
      ~ is_up_to_date               = true -> (known after apply)
      ~ kubeconfig                  = (sensitive value)
        name                        = "management"
      ~ next_upgrade_versions       = [] -> (known after apply)
      ~ nodes_url                   = "REMOVED" -> (known after apply)
      ~ status                      = "READY" -> (known after apply)
      ~ update_policy               = "ALWAYS_UPDATE" -> (known after apply)
      ~ url                         = "REMOVED" -> (known after apply)
      - version                     = "1.20" -> null # forces replacement
        # (3 unchanged attributes hidden)
    }

  # ovh_cloud_project_kube_nodepool.pool must be replaced
-/+ resource "ovh_cloud_project_kube_nodepool" "pool" {
      ~ available_nodes  = 3 -> (known after apply)
      ~ created_at       = "2021-06-23T08:52:45Z" -> (known after apply)
      ~ current_nodes    = 3 -> (known after apply)
      ~ flavor           = "b2-7" -> (known after apply)
      ~ id               = "REMOVED" -> (known after apply)
      ~ kube_id          = "REMOVED" -> (known after apply) # forces replacement
        name             = "pool001"
      ~ project_id       = "REMOVED" -> (known after apply)
      ~ size_status      = "CAPACITY_OK" -> (known after apply)
      ~ status           = "READY" -> (known after apply)
      ~ up_to_date_nodes = 3 -> (known after apply)
      ~ updated_at       = "2021-06-23T08:56:18Z" -> (known after apply)
        # (8 unchanged attributes hidden)
    }

Steps to Reproduce

Please list the steps required to reproduce the issue, for example:

1. terraform apply
2. terraform apply

References

Are there any other GitHub issues (open or closed) or Pull Requests that should be linked here?

• ovh_cloud_project_kube version not optional #183

[yanndegat]

hi @0xErnie

thanks for reporting this. this is strange as we have released a fix in v0.13.0 with a proper functional test regarding this scenario.

we're going to investigate further.

meanwhile, you can use the lifecycle ... ignore_changes = [version] trick

regards

[yanndegat]

hi @0xErnie

FYI, i've logged an issue on the terraform repo. (hashicorp/terraform#29028)
Other issues have already been logged regarding the notable change introduced by terraform 0.15.4
(see: https://github.com/hashicorp/terraform/blob/v0.15/CHANGELOG.md)

AFAIK, they are still open issues, so i would suggest to stick to terraform 0.15.3 for the moment.

hashicorp/terraform#28776
hashicorp/terraform#28803

[yanndegat]

hi @0xErnie

i finally found the root cause of my issue. explaining why i've been misleaded several times.
there's been an issue during the release of the v0.13.0 version of the provider.

it has been released 2 times:

once without the patch fixing the version, once with it. But the second release failed to update the registry and overwrite the first one.
so the v0.13.0 tag on the repository is not reflecting the produced binary on the registry.

you should either pull the v0.13.1 version which properly includes the patchset or upgrade to v0.14.0

regards

is this ok for you?

[yanndegat]

hi @0xErnie

FYI, i've logged an issue on the terraform repo. (hashicorp/terraform#29028)
Other issues have already been logged regarding the notable change introduced by terraform 0.15.4
(see: https://github.com/hashicorp/terraform/blob/v0.15/CHANGELOG.md)

AFAIK, they are still open issues, so i would suggest to stick to terraform 0.15.3 for the moment.

hashicorp/terraform#28776
hashicorp/terraform#28803

this is a wrong analysis

[0xErnie]

Can not verify the soluton since we decided to stop using managed k8s but this seems to be fixed.
Thank you!