What causes "linode subnets %q and %q normalize to the same label %q" in kubernetes/kops?

Rename the subnets with clearly distinct, short, lowercase hyphenated names (e.g. us-east-1a, us-east-1b).

linode subnets %q and %q normalize to the same label %q

Error message

linode subnets %q and %q normalize to the same label %q

What it means

Linode resource labels have strict character rules, so kOps normalizes each subnet label as NormalizeLinodeLabel(clusterName + "-" + subnetName). If two distinct subnets normalize to the same label, tasks would collide and the build aborts with this error identifying the two subnet names and the shared label.

Source

Thrown at pkg/model/linodemodel/network.go:60

	for i, subnet := range b.Cluster.Spec.Networking.Subnets {
		subnetName := subnet.Name
		if subnetName == "" {
			subnetName = fmt.Sprintf("subnet %d", i)
		}

		if subnet.Name == "" {
			return fmt.Errorf("linode subnet %q requires a name", subnetName)
		}
		if subnet.Region == "" {
			return fmt.Errorf("linode subnet %q requires a region", subnetName)
		}
		if subnet.CIDR == "" {
			return fmt.Errorf("linode subnet %q requires a CIDR", subnetName)
		}

		normalizedSubnetName := linode.NormalizeLinodeLabel(b.ClusterName() + "-" + subnet.Name)
		if previousSubnetName, found := seenSubnetNames[normalizedSubnetName]; found {
			return fmt.Errorf("linode subnets %q and %q normalize to the same label %q", previousSubnetName, subnetName, normalizedSubnetName)
		}
		seenSubnetNames[normalizedSubnetName] = subnetName

		if region == "" {
			region = subnet.Region
			continue
		}
		if subnet.Region != region {
			return fmt.Errorf("linode subnets must all use the same region; found %q and %q", region, subnet.Region)
		}
	}

	name := linode.NormalizeLinodeLabel(b.ClusterName())
	description := fmt.Sprintf("kOps VPC for %s", b.ClusterName())
	vpcTask := &linodetasks.VPC{
		Name:        new(name),
		Lifecycle:   b.Lifecycle,
		Description: new(description),

View on GitHub (pinned to 4c8573c808)

Solutions

Rename the subnets with clearly distinct, short, lowercase hyphenated names (e.g. us-east-1a, us-east-1b).
Shorten the cluster name or subnet names so normalized labels stay within Linode's 64-char limit without truncation collisions.
Precompute NormalizeLinodeLabel(clusterName + "-" + subnetName) for all subnets and confirm uniqueness before applying.
Update the InstanceGroups to reference the renamed subnets and re-run kops update cluster.

Example fix

// before
subnets:
- name: us_east_1a
- name: us-east-1a   # normalizes to the same label
// after
subnets:
- name: us-east-1a
- name: us-east-1b

Defensive patterns

Strategy: validation

Validate before calling

seen := map[string]string{}
for _, s := range cluster.Spec.Networking.Subnets {
    label := linode.NormalizeLinodeLabel(clusterName + "-" + s.Name)
    if prev, ok := seen[label]; ok {
        return fmt.Errorf("subnets %q and %q collide on label %q", prev, s.Name, label)
    }
    seen[label] = s.Name
}

Prevention

Choose short, distinct, lowercase-hyphenated subnet names.
Keep cluster name short enough that normalized labels fit Linode's 64-char limit.
Pre-normalize and check uniqueness of all resource labels before apply.
Avoid underscore/dot variants of the same name in one cluster spec.

When it happens

Trigger: Two cluster.spec.networking.subnets entries whose names differ only in characters stripped/replaced by Linode label normalization (e.g. "us_east_1a" vs "us-east-1a"), or names so long that truncation to Linode's label limit makes them identical.

Common situations: Subnet names using underscores/dots/uppercase that normalize to the same hyphenated form; very long cluster+subnet names truncated identically; duplicated subnet entries with trivially different names.

Related errors

AI-assisted analysis of kubernetes/kops@4c8573c808 (2026-09-05). Data as JSON: /api/errors/e9a3840891a9e2e7. Report an issue: GitHub.