Turtles reporting ProviderUnknown for known upstream providers

[cpinjani]

What steps did you take and what happened?

Steps:

• Install community Rancher v2.14.2-alpha6, Turtles v0.26.2-rc.3
• Install any supported providers eg. aws, using the provider charts
• Inspect the provider status > conditions

apiVersion: v1
items:
- apiVersion: turtles-capi.cattle.io/v1alpha1
  kind: CAPIProvider
  metadata:
    annotations:
      meta.helm.sh/release-name: rancher-turtles-providers
      meta.helm.sh/release-namespace: cattle-turtles-system
      operator.cluster.x-k8s.io/applied-spec-hash: 95b4c8953bdfbbbdae7ad85b691c636e9201037f32b846ff0cf947a4088c5ca6
    creationTimestamp: "2026-05-21T18:13:50Z"
    finalizers:
    - provider.cluster.x-k8s.io
    generation: 2
    labels:
      app.kubernetes.io/managed-by: Helm
    name: aws
    namespace: capa-system
    resourceVersion: "7509"
    uid: cfecaa7b-d870-4385-b9ce-08eead1a5465
  spec:
    enableAutomaticUpdate: true
    name: aws
    type: infrastructure
    variables:
      AWS_B64ENCODED_CREDENTIALS: ""
    version: v2.11.1
  status:
    conditions:
    - lastTransitionTime: "2026-05-21T18:14:14Z"
      message: Deployment has minimum availability.
      observedGeneration: 2
      reason: MinimumReplicasAvailable
      status: "True"
      type: Ready
    - lastTransitionTime: "2026-05-21T18:14:05Z"
      message: Provider is unknown
      observedGeneration: 2
      reason: ProviderUnknown
      status: Unknown
      type: CheckLatestVersionTime
    - lastTransitionTime: "2026-05-21T18:14:05Z"
      message: All preflight checks passed
      observedGeneration: 1
      reason: PreflightChecksPassed
      status: "True"
      type: PreflightCheckPassed
    - lastTransitionTime: "2026-05-21T18:14:08Z"
      message: Provider installed successfully
      observedGeneration: 1
      reason: ProviderInstalled
      status: "True"
      type: ProviderInstalled
    - lastTransitionTime: "2026-05-21T18:14:05Z"
      message: Certificates are now managed by wrangler
      observedGeneration: 1
      reason: CertificatesManaged
      status: "True"
      type: WranglerManagedCertificates
    contract: v1beta1
    installedVersion: v2.11.1
    name: aws
    observedGeneration: 2
    phase: Ready
    variables:
      CLUSTER_TOPOLOGY: "true"
      EXP_CLUSTER_RESOURCE_SET: "true"
      EXP_MACHINE_POOL: "true"
kind: List
metadata:
  resourceVersion: ""

What did you expect to happen?

Turtles should lookup the upstream clusterctl config when a provider is not found, which is equivalent of running clusterctl config repositories, and not report provider as unknown.
https://turtles.docs.rancher.com/turtles/stable/en/reference/capiprovider.html#_version_pinning_and_automatic_updates

How to reproduce it?

No response

Rancher Turtles version

No response

Anything else you would like to add?

No response

Label(s) to be applied

/kind bug

[anmazzotti]

Mainly the problem is the inconsistency of the "latest" version calculation, with the behavior of cluster-api-operator.
Turtles only checks the installed clusterctlconfig in /config/clusterctl.yaml, but cluster-api-operator also adds a list of upstream overrides.

See setLatestVersion function for turtles.
See usage of client.Providers.List() in cluster-api-operator.

This problem was introduced by this refactor: #1601

The fix is going to be hard because we need to cope with the fact that we need CAPIProvider.spec.version to be initially empty to work with the hardcoded values. On top of that CAPIProvider.spec.enableAutomaticUpdates was also needed to cope with the same hardcoded version mechanism and allow providers to automatically bump via hardcoded values when a new version of Turtles is installed, which only makes things complicated.

EnabledAutomaticUpdates should be deprecated and the feature removed because it's not relevant to users and does not align to clusterctl either.

So overall I think a good structural fix is blocked by #2041

[anmazzotti]

By the way a workaround is to use ClusterctlConfig.
This should work for Community users to have the provider actually known.

https://turtles.docs.rancher.com/turtles/stable/en/reference/clusterctlconfig.html

[anmazzotti]

Putting it back in progress as I'm still trying to find a fix in #2453

[anmazzotti]

Nope, no quick fix I can come up with.