kubernetes/kops · error

unable to determine gcp network for cluster

Error message

unable to determine gcp network for cluster

What it means

applyGCPCluster falls back to deriving the network name from the cluster name when spec.networkID is empty; if even that fallback yields an empty string it fails with this error. The GCPCluster object needs a GCP network reference. In practice this is nearly unreachable, since SafeTruncatedClusterName derives from ObjectMeta.Name.

Source

Thrown at pkg/controllers/clusterapi/cluster_controller.go:137

	gcpRegion := ""
	for _, subnet := range s.Cluster.Spec.Subnets {
		if gcpRegion == "" {
			gcpRegion = subnet.Region
		} else if gcpRegion != subnet.Region {
			return fmt.Errorf("found multiple gcp regions for cluster")
		}
	}
	if gcpRegion == "" {
		return fmt.Errorf("unable to determine gcp region for cluster")
	}

	// TODO: Sync with LinkToNetwork
	gcpNetworkName := s.Cluster.Spec.NetworkID
	if gcpNetworkName == "" {
		gcpNetworkName = gce.SafeTruncatedClusterName(s.Cluster.ObjectMeta.Name, 63)
	}
	if gcpNetworkName == "" {
		return fmt.Errorf("unable to determine gcp network for cluster")
	}

	obj := map[string]any{
		"apiVersion": "infrastructure.cluster.x-k8s.io/v1beta1",
		"kind":       "GCPCluster",
		"metadata": map[string]any{
			"name":      name,
			"namespace": s.namespace(),
		},
		"spec": map[string]any{
			"project": gcpProject,
			"region":  gcpRegion,
			"network": map[string]any{
				"name": gcpNetworkName,
			},
		},
	}

View on GitHub (pinned to 4c8573c808)

Solutions

1. Set spec.networkID to the existing GCP VPC network name
2. Ensure Cluster metadata.name is populated
3. Re-run reconciliation after fixing the object

Example fix

// before
spec:
  subnets: [...]   # networkID missing
// after
spec:
  networkID: my-vpc-network
  subnets: [...]

if cluster.Spec.NetworkID == "" {
	return fmt.Errorf("cluster %s: spec.networkID recommended; fallback only works if name is set", cluster.Name)
}

When it happens

Trigger: clusterScope.Apply() where spec.networkID == '' and SafeTruncatedClusterName(cluster name) returns '' — effectively only when the Cluster has an empty metadata.name.

Common situations: Programmatically constructed Cluster objects with missing metadata.name; test fixtures; manifests corrupted before reconciliation.

Related errors

• unable to determine gcp project for cluster
• found multiple gcp regions for cluster
• unable to determine gcp region for cluster
• applying GCPCluster object to cluster: %w
• error fetching instance from compute API: %w

AI-assisted analysis of kubernetes/kops@4c8573c808 (2026-09-05). Data as JSON: /api/errors/4abda5132d4d1887. Report an issue: GitHub.