no instance types were returned because the criteria specifi

Error message

no instance types were returned because the criteria specified was too narrow

What it means

Filter() succeeded but returned zero instance types, meaning the supplied criteria (vCPU:memory ratio, flexible window, usage class, zones) match no EC2 instance type. kOps treats an empty result as a user-input error rather than a system failure.

Source

Thrown at cmd/kops/toolbox_instance-selector.go:300

		if filters.VCpusToMemoryRatio == nil {
			defaultStartRatio := float64(2.0)
			mutatedFilters.VCpusToMemoryRatio = &defaultStartRatio
		}
	}

	newInstanceGroups := []*kops.InstanceGroup{}

	for i := 0; i < igCount; i++ {
		igNameForRun := options.InstanceGroupName
		if igCount != 1 {
			igNameForRun = fmt.Sprintf("%s%d", options.InstanceGroupName, i+1)
		}
		selectedInstanceTypes, err := instanceSelector.Filter(ctx, mutatedFilters)
		if err != nil {
			return fmt.Errorf("error finding matching instance types: %w", err)
		}
		if len(selectedInstanceTypes) == 0 {
			return fmt.Errorf("no instance types were returned because the criteria specified was too narrow")
		}
		usageClass := *filters.UsageClass

		ig := createInstanceGroup(igNameForRun, options.ClusterName, igSubnets)
		ig = decorateWithInstanceGroupSpecs(ig, options)
		ig, err = decorateWithMixedInstancesPolicy(ig, usageClass, selectedInstanceTypes)
		if err != nil {
			return err
		}
		if options.ClusterAutoscaler {
			ig = decorateWithClusterAutoscalerLabels(ig, options.ClusterName)
		}
		ig, err = cloudup.PopulateInstanceGroupSpec(cluster, ig, cloud, channel)
		if err != nil {
			return err
		}

		newInstanceGroups = append(newInstanceGroups, ig)

View on GitHub (pinned to 4c8573c808)

Solutions

Widen the --vcpus-to-memory-ratio range or remove it entirely.
Increase --flexible so more candidate shapes can qualify.
Allow more zones (more subnets) instead of one AZ.
Switch usage class from spot to regular, or vice versa, if that is the binding constraint.

Example fix

// before
kops toolbox instance-selector --vcpus 4 --memory 4 --flexible 1 --zones us-east-1a
// after
kops toolbox instance-selector --vcpus 2 --memory 8 --flexible 5

Defensive patterns

Strategy: fallback

Validate before calling

func filtersTooNarrow(f selector.Filters) bool {
	return (f.Flexible != nil && *f.Flexible < 2) ||
		(f.VCpusToMemoryRatio != nil && (*f.VCpusToMemoryRatio < 1 || *f.VCpusToMemoryRatio > 16))
}

Try / catch

selected, err := instanceSelector.Filter(ctx, mutatedFilters)
if err == nil && len(selected) == 0 {
	mutatedFilters.Flexible = ptr.To(10) // widen and retry
	selected, err = instanceSelector.Filter(ctx, mutatedFilters)
}

Prevention

Prefer omitting --vcpus-to-memory-ratio unless you need it.
Use --flexible >= 3 so multiple shapes can match.
Target multiple zones instead of a single AZ.

When it happens

Trigger: instanceSelector.Filter returns a non-empty-nil slice (len == 0) for the mutated filters — e.g. a very narrow vcpus-to-memory ratio with small flexible bounds, or spot-only requests in a zone with no matching families.

Common situations: Specifying an extreme ratio like 1:16 or 16:1 with --flexible 1; targeting a single zone with limited families; requesting spot capacity with tight constraints during spot shortages.

Related errors

AI-assisted analysis of kubernetes/kops@4c8573c808 (2026-09-05). Data as JSON: /api/errors/d9582d3ec04eb96c. Report an issue: GitHub.