k3s-io/k3s · error

all cloud-provider functionality disabled by config

Error message

all cloud-provider functionality disabled by config

What it means

The k3s cloud provider registers under the name 'k3s' with lbEnabled and nodeEnabled defaulting to true. If the JSON cloud-config supplied to the controller-manager sets both to false, the provider has nothing to run and registration returns an error instead of a silent no-op provider.

Source

Thrown at pkg/cloudprovider/cloudprovider.go:78

			Config: Config{
				LBDefaultPriorityClassName: DefaultLBPriorityClassName,
				LBEnabled:                  true,
				LBImage:                    DefaultLBImage,
				LBNamespace:                DefaultLBNS,
				NodeEnabled:                true,
			},
		}

		if config != nil {
			var bytes []byte
			bytes, err = io.ReadAll(config)
			if err == nil {
				err = json.Unmarshal(bytes, &k.Config)
			}
		}

		if !k.LBEnabled && !k.NodeEnabled {
			return nil, errors.New("all cloud-provider functionality disabled by config")
		}

		return &k, err
	})
}

func (k *k3s) Initialize(clientBuilder cloudprovider.ControllerClientBuilder, stop <-chan struct{}) {
	ctx := logger.NewContext(wait.ContextForChannel(stop), controllerName)
	config := clientBuilder.ConfigOrDie(controllerName)
	k.client = kubernetes.NewForConfigOrDie(config)

	if k.LBEnabled {
		// Wrangler controller and caches are only needed if the load balancer controller is enabled.
		k.recorder = util.BuildControllerEventRecorder(ctx, k.client, controllerName, meta.NamespaceAll)
		coreFactory := core.NewFactoryFromConfigOrDie(config)
		k.nodeCache = coreFactory.Core().V1().Node().Cache()

		lbCoreFactory := core.NewFactoryFromConfigWithOptionsOrDie(config, &generic.FactoryOptions{Namespace: k.LBNamespace})

View on GitHub (pinned to 6ba341e396)

Solutions

1. If neither function is wanted, do not load the k3s cloud provider at all — drop --cloud-provider=k3s / the cloud-config file
2. Keep at least one of lbEnabled/nodeEnabled true in the JSON config
3. To disable only ServiceLB, use the k3s flag --disable-service-lb rather than blanking cloud-config

Example fix

# before (cloud-config.json)
{"lbEnabled":false,"nodeEnabled":false}

# after (ServiceLB off, node controller still active)
{"lbEnabled":false,"nodeEnabled":true}

// validate cloud-config before handing it to the controller-manager
type k3sCCMConfig struct {
    LBEnabled   *bool `json:"lbEnabled"`
    NodeEnabled *bool `json:"nodeEnabled"`
}
var c k3sCCMConfig
if err := json.Unmarshal(raw, &c); err == nil {
    lb := c.LBEnabled == nil || *c.LBEnabled
    node := c.NodeEnabled == nil || *c.NodeEnabled
    if !lb && !node {
        return errors.New("cloud-config disables all k3s provider functionality; omit --cloud-provider=k3s instead")
    }
}

func k3sProviderConfigUsable(cfg []byte) bool {
    var c struct {
        LBEnabled   *bool `json:"lbEnabled"`
        NodeEnabled *bool `json:"nodeEnabled"`
    }
    if json.Unmarshal(cfg, &c) != nil {
        return true // parse error handled elsewhere
    }
    lb := c.LBEnabled == nil || *c.LBEnabled
    node := c.NodeEnabled == nil || *c.NodeEnabled
    return lb || node
}

When it happens

Trigger: Starting kube-controller-manager / k3s's embedded cloud controller with `--cloud-provider=k3s --cloud-config=<file>` where the file is `{"lbEnabled":false,"nodeEnabled":false}`.

Common situations: Operators trying to switch off ServiceLB and the node controller via cloud-config while still loading the k3s provider; config files copied from docs with both toggles flipped; leftover cloud-config when migrating to a different CCM.

Related errors

• etcd-s3-timeout must be greater than 0s
• toleration with empty key must have operator 'Exists'
• toleration with operator 'Exists' must have an empty value
• critical configuration value mismatch between servers
• s3 bucket name was not set

AI-assisted analysis of k3s-io/k3s@6ba341e396 (2026-08-15). Data as JSON: /api/errors/bc012d7c183fb0e5. Report an issue: GitHub.