What causes "invalid value for MaxUnavailable: %v" in kubernetes/kubernetes?

Inspect the DS: kubectl get ds -o yaml and check spec.updateStrategy.rollingUpdate.maxUnavailable.

invalid value for MaxUnavailable: %v

Error message

invalid value for MaxUnavailable: %v

What it means

Wraps a failure from util.UnavailableCount (daemonset_util.go:153) during updatedDesiredNodeCounts (update.go:609). UnavailableCount only errors when GetScaledValueFromIntOrPercent cannot parse the RollingUpdate.MaxUnavailable IntOrString (e.g. a malformed percentage string, empty, or a value that overflows when scaled). For a RollingUpdate DaemonSet this blocks rollout math, so the whole reconcile for that DS fails and requeues.

Source

Thrown at pkg/controller/daemon/update.go:609

	logger := klog.FromContext(ctx)
	tolerations := daemonPodTolerations(ds)
	requiredNodeAffinity := nodeaffinity.NewRequiredNodeAffinity(ds.Spec.Template.Spec.NodeSelector, ds.Spec.Template.Spec.Affinity)
	for i := range nodeList {
		node := nodeList[i]
		wantToRun, _ := nodeShouldRunDaemonPod(logger, node, ds, tolerations, requiredNodeAffinity)
		if !wantToRun {
			continue
		}
		desiredNumberScheduled++

		if _, exists := nodeToDaemonPods[node.Name]; !exists {
			nodeToDaemonPods[node.Name] = nil
		}
	}

	maxUnavailable, err := util.UnavailableCount(ds, desiredNumberScheduled)
	if err != nil {
		return -1, -1, -1, fmt.Errorf("invalid value for MaxUnavailable: %v", err)
	}

	maxSurge, err := util.SurgeCount(ds, desiredNumberScheduled)
	if err != nil {
		return -1, -1, -1, fmt.Errorf("invalid value for MaxSurge: %v", err)
	}

	// if the daemonset returned with an impossible configuration, obey the default of unavailable=1 (in the
	// event the apiserver returns 0 for both surge and unavailability)
	if desiredNumberScheduled > 0 && maxUnavailable == 0 && maxSurge == 0 {
		logger.Info("DaemonSet is not configured for surge or unavailability, defaulting to accepting unavailability", "daemonset", klog.KObj(ds))
		maxUnavailable = 1
	}
	logger.V(5).Info("DaemonSet with maxSurge and maxUnavailable", "daemonset", klog.KObj(ds), "maxSurge", maxSurge, "maxUnavailable", maxUnavailable)
	return maxSurge, maxUnavailable, desiredNumberScheduled, nil
}

type historiesByRevision []*apps.ControllerRevision

View on GitHub (pinned to b882c60b40)

Solutions

Inspect the DS: kubectl get ds <name> -o yaml and check spec.updateStrategy.rollingUpdate.maxUnavailable.
Set maxUnavailable to a valid int (e.g. 1) or percent string (e.g. "25%") and kubectl apply.
If generated by Helm/kustomize, fix the template so the field renders a concrete intstr value.

Example fix

// before
spec:
  updateStrategy:
    type: RollingUpdate
    rollingUpdate:
      maxUnavailable: ""   # unparseable -> error
// after
spec:
  updateStrategy:
    type: RollingUpdate
    rollingUpdate:
      maxUnavailable: 1

Defensive patterns

Strategy: validation

Validate before calling

// Validate maxUnavailable before relying on it.
func validMaxUnavailable(v *intstr.IntOrString) bool {
    if v == nil { return true }
    _, err := intstrutil.GetScaledValueFromIntOrPercent(v, 1, true)
    return err == nil
}

Try / catch

if _, _, _, err := dsc.updatedDesiredNodeCounts(ctx, ds, nodes, nodeToPods); err != nil {
    if strings.Contains(err.Error(), "invalid value for MaxUnavailable") {
        // Bad user input; surface as a Deployment/DaemonSet event, do not hot-loop.
        dsc.eventRecorder.Eventf(ds, v1.EventTypeWarning, "InvalidMaxUnavailable", err.Error())
        return reconcile.Result{RequeueAfter: 30 * time.Second}, nil
    }
    return reconcile.Result{}, err
}

Prevention

Run kubectl apply --dry-run=server to let admission validate intstr fields.
Render manifests through templates and assert maxUnavailable is a concrete int/percent in CI.
Default maxUnavailable explicitly rather than relying on an empty/ambiguous value.

When it happens

Trigger: DS has spec.updateStrategy.type=RollingUpdate but rollingUpdate.maxUnavailable is set to an unparseable value (e.g. "-1%", "abc", "") or an intstr that GetScaledValueFromIntOrPercent rejects for the given desiredNumberScheduled.

Common situations: Hand-edited YAML with a typo in maxUnavailable; a templating/Helm rendering bug emitting an empty field; API version mismatch where the field shape changed.

Related errors

AI-assisted analysis of kubernetes/kubernetes@b882c60b40 (2026-08-07). Data as JSON: /api/errors/14225f30c1ede3af. Report an issue: GitHub.