Merge pull request #10430 from k8s-infra-cherrypick-robot/cherry-pick…

…-10087-to-release-1.7 [release-1.7] 🐛 fix: deletion priority to avoid deleting too many machines
kubernetes-sigs · Apr 15, 2024 · ce58362 · ce58362
2 parents f4cdfac + b2b4f13
commit ce58362
Show file tree

Hide file tree

Showing 2 changed files with 36 additions and 6 deletions.
diff --git a/internal/controllers/machineset/machineset_delete_policy.go b/internal/controllers/machineset/machineset_delete_policy.go
@@ -35,6 +35,7 @@ type (
 
 const (
 	mustDelete    deletePriority = 100.0
+	shouldDelete  deletePriority = 75.0
 	betterDelete  deletePriority = 50.0
 	couldDelete   deletePriority = 20.0
 	mustNotDelete deletePriority = 0.0
@@ -48,10 +49,10 @@ func oldestDeletePriority(machine *clusterv1.Machine) deletePriority {
 		return mustDelete
 	}
 	if _, ok := machine.ObjectMeta.Annotations[clusterv1.DeleteMachineAnnotation]; ok {
-		return mustDelete
+		return shouldDelete
 	}
 	if !isMachineHealthy(machine) {
-		return mustDelete
+		return betterDelete
 	}
 	if machine.ObjectMeta.CreationTimestamp.Time.IsZero() {
 		return mustNotDelete
@@ -60,20 +61,20 @@ func oldestDeletePriority(machine *clusterv1.Machine) deletePriority {
 	if d.Seconds() < 0 {
 		return mustNotDelete
 	}
-	return deletePriority(float64(mustDelete) * (1.0 - math.Exp(-d.Seconds()/secondsPerTenDays)))
+	return deletePriority(float64(betterDelete) * (1.0 - math.Exp(-d.Seconds()/secondsPerTenDays)))
 }
 
 func newestDeletePriority(machine *clusterv1.Machine) deletePriority {
 	if !machine.DeletionTimestamp.IsZero() {
 		return mustDelete
 	}
 	if _, ok := machine.ObjectMeta.Annotations[clusterv1.DeleteMachineAnnotation]; ok {
-		return mustDelete
+		return shouldDelete
 	}
 	if !isMachineHealthy(machine) {
-		return mustDelete
+		return betterDelete
 	}
-	return mustDelete - oldestDeletePriority(machine)
+	return betterDelete - oldestDeletePriority(machine)
 }
 
 func randomDeletePolicy(machine *clusterv1.Machine) deletePriority {

diff --git a/internal/controllers/machineset/machineset_delete_policy_test.go b/internal/controllers/machineset/machineset_delete_policy_test.go
@@ -407,6 +407,18 @@ func TestMachineOldestDelete(t *testing.T) {
 		ObjectMeta: metav1.ObjectMeta{CreationTimestamp: metav1.NewTime(currentTime.Time.AddDate(0, 0, -10))},
 		Status:     clusterv1.MachineStatus{FailureReason: &statusError, NodeRef: nodeRef},
 	}
+	mustDeleteMachine := &clusterv1.Machine{
+		ObjectMeta: metav1.ObjectMeta{Name: "b", DeletionTimestamp: &currentTime},
+		Status:     clusterv1.MachineStatus{NodeRef: nodeRef},
+	}
+	unhealthyMachineA := &clusterv1.Machine{
+		ObjectMeta: metav1.ObjectMeta{Name: "a", CreationTimestamp: metav1.NewTime(currentTime.Time.AddDate(0, 0, -10))},
+		Status:     clusterv1.MachineStatus{FailureReason: &statusError, NodeRef: nodeRef},
+	}
+	unhealthyMachineZ := &clusterv1.Machine{
+		ObjectMeta: metav1.ObjectMeta{Name: "z", CreationTimestamp: metav1.NewTime(currentTime.Time.AddDate(0, 0, -10))},
+		Status:     clusterv1.MachineStatus{FailureReason: &statusError, NodeRef: nodeRef},
+	}
 	deleteMachineWithoutNodeRef := &clusterv1.Machine{
 		ObjectMeta: metav1.ObjectMeta{CreationTimestamp: metav1.NewTime(currentTime.Time.AddDate(0, 0, -10))},
 	}
@@ -513,6 +525,23 @@ func TestMachineOldestDelete(t *testing.T) {
 			},
 			expect: []*clusterv1.Machine{nodeHealthyConditionUnknownMachine},
 		},
+		// these two cases ensures the mustDeleteMachine is always picked regardless of the machine names.
+		{
+			desc: "func=oldestDeletePriority, diff=1 (unhealthyMachineA)",
+			diff: 1,
+			machines: []*clusterv1.Machine{
+				empty, secondNewest, oldest, secondOldest, newest, mustDeleteMachine, unhealthyMachineA,
+			},
+			expect: []*clusterv1.Machine{mustDeleteMachine},
+		},
+		{
+			desc: "func=oldestDeletePriority, diff=1 (unhealthyMachineZ)",
+			diff: 1,
+			machines: []*clusterv1.Machine{
+				empty, secondNewest, oldest, secondOldest, newest, mustDeleteMachine, unhealthyMachineZ,
+			},
+			expect: []*clusterv1.Machine{mustDeleteMachine},
+		},
 	}
 
 	for _, test := range tests {