`cluster_api_url` format change forces resource replacement during provider upgrade

[antoinedeschenes]

When upgrading from v0.2.5 to v1.6.0 of the Redpanda Terraform provider, the cluster_api_url attribute format has changed across all resources that use it, causing all affected resources to be marked for replacement.

I believe the provider should do a state migration to move to the new URL format without triggering a resource replacement.

-/+ resource "redpanda_acl" "groups" {
+       allow_deletion        = false
!       cluster_api_url       = "api-xxxxx.example.byoc.prd.cloud.redpanda.com:443" -> "https://api-xxxxx.example.byoc.prd.cloud.redpanda.com" # forces replacement
!       id                    = "GROUP:*:LITERAL:User:app:*:ALL:ALLOW" -> (known after apply)
        # (7 unchanged attributes hidden)
    }

-/+ resource "redpanda_acl" "topics" {
+       allow_deletion        = false
!       cluster_api_url       = "api-xxxxx.example.byoc.prd.cloud.redpanda.com:443" -> "https://api-xxxxx.example.byoc.prd.cloud.redpanda.com" # forces replacement
!       id                    = "TOPIC:*:LITERAL:User:app:*:ALL:ALLOW" -> (known after apply)
        # (7 unchanged attributes hidden)
    }

-/+ resource "redpanda_acl" "transactions" {
+       allow_deletion        = false
!       cluster_api_url       = "api-xxxxx.example.byoc.prd.cloud.redpanda.com:443" -> "https://api-xxxxx.example.byoc.prd.cloud.redpanda.com" # forces replacement
!       id                    = "TRANSACTIONAL_ID:*:LITERAL:User:app:*:ALL:ALLOW" -> (known after apply)
        # (7 unchanged attributes hidden)
    }

-/+ resource "redpanda_user" "user" {
+       allow_deletion  = false
!       cluster_api_url = "api-xxxxx.example.byoc.prd.cloud.redpanda.com:443" -> "https://api-xxxxx.example.byoc.prd.cloud.redpanda.com" # forces replacement
!       id              = "app" -> (known after apply)
        name            = "app"
        # (3 unchanged attributes hidden)
    }

-/+ resource "redpanda_topic" "topics" {
+       allow_deletion     = false
!       cluster_api_url    = "api-xxxxx.example.byoc.prd.cloud.redpanda.com:443" -> "https://api-xxxxx.example.byoc.prd.cloud.redpanda.com" # forces replacement
!       id                 = "app" -> (known after apply)
        name               = "app"
        # (3 unchanged attributes hidden)
    }

[david-yu]

Thank you for the feedback. We do want to handle such migrations in the future however the current workaround is to remove and reimport the rsources with Import State (i.e. users and topics).

You would need to do something like the following:

  # Find your cluster ID from state or the Redpanda Cloud console
  CLUSTER_ID="<your-cluster-id>"

  # Remove resources from state
  terraform state rm redpanda_user.user
  terraform state rm redpanda_topic.topics

  # Reimport — the ImportState implementation looks up cluster_api_url from
  # the control plane using the cluster ID, so it will store the new URL format
  terraform import redpanda_user.user "app,${CLUSTER_ID}"
  terraform import redpanda_topic.topics "app,${CLUSTER_ID}"

  For the three ACLs, you cannot use terraform import — redpanda_acl doesn't implement ImportState. Use Option A for ACLs, or:

  # Remove ACLs from state, then re-apply
  # CreateACL is typically idempotent (the ACLs already exist on the cluster)
  terraform state rm 'redpanda_acl.groups'
  terraform state rm 'redpanda_acl.topics'
  terraform state rm 'redpanda_acl.transactions'
  terraform apply   # Will call CreateACL which should succeed even if they exist

[antoinedeschenes]

Thanks @david-yu we'll consider this. Do you think we could also consider patching the state with a pull/push?