Migrating SKS List Data Sources from v0.72.x to v0.73.x

This guide covers the migration of the exoscale_sks_cluster_list and exoscale_sks_nodepool_list data sources from provider version ~> 0.72.x to ~> 0.73.x.

Overview

Version 0.73.0 migrates exoscale_sks_cluster_list and exoscale_sks_nodepool_list from the legacy SDKv2 implementation to the Terraform plugin framework, continuing the SKS migration started with exoscale_sks_cluster, exoscale_sks_nodepool and exoscale_sks_kubeconfig in a previous release.

As part of this migration, both data sources drop their generic, per-attribute filter arguments: zone is now the only input they accept, and they always return every cluster / node pool found in that zone. If you relied on server-side filtering, filter the returned list yourself with a Terraform for expression instead. A handful of attributes on the returned list items were also removed, and a few on exoscale_sks_nodepool_list are now populated correctly for the first time — see below.

Since these are data sources, there is no resource state to carry forward or upgrade: they are fully re-read on every plan/apply. The breaking change is entirely in your configuration — the arguments you pass in, and the attributes you reference from the result — not in any stored state.

~> Note: Before migrating your configuration you need to ensure you use the latest version of Terraform and have a clean configuration.

What has Changed

Filtering is no longer built in

Previously, every attribute of a cluster/node pool doubled as an optional top-level filter argument on the list data source (with regex support via /.../). That’s gone: both data sources now only accept zone (and timeouts) as input, and always return the full list for that zone.

Before (v0.72.x):

data "exoscale_sks_cluster_list" "prod" {
  zone = "de-fra-1"
  labels = {
    "customer" = "/.*telecom.*/"
  }
}

After (v0.73.x):

data "exoscale_sks_cluster_list" "all" {
  zone = "de-fra-1"
}

locals {
  prod_clusters = [
    for c in data.exoscale_sks_cluster_list.all.clusters :
    c if can(regex(".*telecom.*", lookup(c.labels, "customer", "")))
  ]
}

The same applies to exoscale_sks_nodepool_list:

Before (v0.72.x):

data "exoscale_sks_nodepool_list" "workers" {
  zone = "de-fra-1"
  name = "/.*-workers-.*/"
}

After (v0.73.x):

data "exoscale_sks_nodepool_list" "all" {
  zone = "de-fra-1"
}

locals {
  worker_nodepools = [
    for np in data.exoscale_sks_nodepool_list.all.nodepools :
    np if can(regex(".*-workers-.*", np.name))
  ]
}

Removed attributes on exoscale_sks_cluster_list items

The following attributes have been removed from each entry of clusters: aggregation_ca, audit, control_plane_ca, create_default_security_group, enable_karpenter, exoscale_ccm, exoscale_csi, kubelet_ca, metrics_server, oidc, and the per-item zone (redundant with the data source’s own zone argument). These attributes were declared in the legacy schema but never actually populated by the provider, so they always read as an empty or zero value — no real data is lost. If you need any of them, look up the specific cluster with the exoscale_sks_cluster data source instead.

exoscale_sks_nodepool_list items: one removed, some fixed, one new

The per-item zone attribute was removed, same as above (redundant with the data source’s own zone argument).

storage_lvm, ipv6, template_id and kubelet_image_gc are now populated from each node pool’s actual state. Previously they were declared in the schema but never set by the provider, so they always read as false/empty regardless of the real node pool configuration — if your configuration depended on that always-empty behavior, double check it against the real values.

A new cluster_id attribute is now populated on each item, identifying which exoscale_sks_cluster the node pool belongs to. This existed in the legacy schema too, but only as an input-only filter argument that could never actually match anything (it was never compared against real node pool data), so it silently returned an empty list whenever used. It’s now a real, always-populated output attribute.

Migration Steps

1. Update Provider Version

terraform {
  required_providers {
    exoscale = {
      source  = "exoscale/exoscale"
      version = "~> 0.73.0"
    }
  }
}

2. Replace filter arguments with a for expression

For every exoscale_sks_cluster_list or exoscale_sks_nodepool_list data source that sets an argument other than zone (or timeouts), remove it and instead filter the clusters/nodepools list yourself, as shown above.

3. Remove references to dropped attributes

Search your configuration for the removed clusters[*] attributes listed above (aggregation_ca, audit, control_plane_ca, create_default_security_group, enable_karpenter, exoscale_ccm, exoscale_csi, kubelet_ca, metrics_server, oidc) and the per-item zone on both data sources, and remove them.

4. Verify Changes

After updating your configuration:

1. Run terraform init -upgrade to upgrade the provider.
2. Run terraform plan to verify there are no further changes beyond what you expect from the newly-populated exoscale_sks_nodepool_list attributes.

You should see output similar to:

No changes. Your infrastructure matches the configuration.

Terraform has compared your real infrastructure against your configuration
and found no differences, so no changes are needed.

Additional Resources

• exoscale_sks_cluster_list Data Source
• exoscale_sks_nodepool_list Data Source