Use maxUnavailable for the critical-op PDB to stop idle alert noise

The critical-op PDB is created with minAvailable equal to
numberOfInstances while its selector (critical-operation=true) matches
no pods during normal operation. This leaves status.desiredHealthy at N
and currentHealthy at 0 permanently, so monitoring stacks fire alerts
like kube-prometheus-stack's KubePdbNotEnoughHealthyPods for every idle
cluster (#3020).

maxUnavailable: 0 provides the same protection while a critical
operation is running - no voluntary evictions of labeled pods - but
keeps the budget satisfied (desiredHealthy 0) when nothing matches.
When PDBs are disabled or there are no instances, the budget relaxes to
maxUnavailable 100% instead of minAvailable 0.

Fixes #3020

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Juhani Pelli 2026-07-24 17:52:19 +03:00
parent bffde0f5e7
commit 61e506dea0
2 changed files with 25 additions and 8 deletions

View File

@ -2315,12 +2315,18 @@ func (c *Cluster) generatePrimaryPodDisruptionBudget() *policyv1.PodDisruptionBu
} }
func (c *Cluster) generateCriticalOpPodDisruptionBudget() *policyv1.PodDisruptionBudget { func (c *Cluster) generateCriticalOpPodDisruptionBudget() *policyv1.PodDisruptionBudget {
minAvailable := intstr.FromInt32(c.Spec.NumberOfInstances) // MaxUnavailable: 0 blocks voluntary disruption of any pod carrying the
// critical-operation label, while keeping the budget satisfied when no
// pod matches (status.desiredHealthy stays 0 outside critical
// operations). The previous MinAvailable: N spec left desiredHealthy at
// N with zero matching pods during normal operation, permanently firing
// alerts like kube-prometheus-stack's KubePdbNotEnoughHealthyPods (#3020).
maxUnavailable := intstr.FromInt32(0)
pdbEnabled := c.OpConfig.EnablePodDisruptionBudget pdbEnabled := c.OpConfig.EnablePodDisruptionBudget
// if PodDisruptionBudget is disabled or if there are no DB pods, set the budget to 0. // if PodDisruptionBudget is disabled or if there are no DB pods, allow all disruptions.
if (pdbEnabled != nil && !(*pdbEnabled)) || c.Spec.NumberOfInstances <= 0 { if (pdbEnabled != nil && !(*pdbEnabled)) || c.Spec.NumberOfInstances <= 0 {
minAvailable = intstr.FromInt(0) maxUnavailable = intstr.FromString("100%")
} }
labels := c.labelsSet(false) labels := c.labelsSet(false)
@ -2335,7 +2341,7 @@ func (c *Cluster) generateCriticalOpPodDisruptionBudget() *policyv1.PodDisruptio
OwnerReferences: c.ownerReferences(), OwnerReferences: c.ownerReferences(),
}, },
Spec: policyv1.PodDisruptionBudgetSpec{ Spec: policyv1.PodDisruptionBudgetSpec{
MinAvailable: &minAvailable, MaxUnavailable: &maxUnavailable,
Selector: &metav1.LabelSelector{ Selector: &metav1.LabelSelector{
MatchLabels: labels, MatchLabels: labels,
}, },

View File

@ -2556,6 +2556,17 @@ func TestGeneratePodDisruptionBudget(t *testing.T) {
} }
} }
hasMaxUnavailable := func(expected string) func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error {
return func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error {
actual := podDisruptionBudget.Spec.MaxUnavailable.String()
if actual != expected {
return fmt.Errorf("PodDisruptionBudget MaxUnavailable is incorrect, got %s, expected %s",
actual, expected)
}
return nil
}
}
hasMinAvailable := func(expectedMinAvailable int) func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error { hasMinAvailable := func(expectedMinAvailable int) func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error {
return func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error { return func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error {
actual := podDisruptionBudget.Spec.MinAvailable.IntVal actual := podDisruptionBudget.Spec.MinAvailable.IntVal
@ -2749,7 +2760,7 @@ func TestGeneratePodDisruptionBudget(t *testing.T) {
check: []func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error{ check: []func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error{
testPodDisruptionBudgetOwnerReference, testPodDisruptionBudgetOwnerReference,
hasName("postgres-myapp-database-critical-op-pdb"), hasName("postgres-myapp-database-critical-op-pdb"),
hasMinAvailable(3), hasMaxUnavailable("0"),
testLabelsAndSelectors(false), testLabelsAndSelectors(false),
}, },
}, },
@ -2766,7 +2777,7 @@ func TestGeneratePodDisruptionBudget(t *testing.T) {
check: []func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error{ check: []func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error{
testPodDisruptionBudgetOwnerReference, testPodDisruptionBudgetOwnerReference,
hasName("postgres-myapp-database-critical-op-pdb"), hasName("postgres-myapp-database-critical-op-pdb"),
hasMinAvailable(0), hasMaxUnavailable("100%"),
testLabelsAndSelectors(false), testLabelsAndSelectors(false),
}, },
}, },
@ -2783,7 +2794,7 @@ func TestGeneratePodDisruptionBudget(t *testing.T) {
check: []func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error{ check: []func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error{
testPodDisruptionBudgetOwnerReference, testPodDisruptionBudgetOwnerReference,
hasName("postgres-myapp-database-critical-op-pdb"), hasName("postgres-myapp-database-critical-op-pdb"),
hasMinAvailable(0), hasMaxUnavailable("100%"),
testLabelsAndSelectors(false), testLabelsAndSelectors(false),
}, },
}, },
@ -2800,7 +2811,7 @@ func TestGeneratePodDisruptionBudget(t *testing.T) {
check: []func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error{ check: []func(cluster *Cluster, podDisruptionBudget *policyv1.PodDisruptionBudget) error{
testPodDisruptionBudgetOwnerReference, testPodDisruptionBudgetOwnerReference,
hasName("postgres-myapp-database-critical-op-pdb"), hasName("postgres-myapp-database-critical-op-pdb"),
hasMinAvailable(3), hasMaxUnavailable("0"),
testLabelsAndSelectors(false), testLabelsAndSelectors(false),
}, },
}, },