fix: deletion priority to avoid deleting too many machines

This introduces an additional deletePriority between betterDelete and
mustDelete to avoid a race where multiple machines might be deleted at
the same time.

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 {

internal/controllers/machineset/machineset_delete_policy_test.go

@@ -383,6 +383,10 @@ func TestMachineOldestDelete(t *testing.T) {
 	empty := &clusterv1.Machine{
 		Status: clusterv1.MachineStatus{NodeRef: nodeRef},
 	}
+	mustDeleteMachine := &clusterv1.Machine{
+		ObjectMeta: metav1.ObjectMeta{DeletionTimestamp: &currentTime},
+		Status:     clusterv1.MachineStatus{NodeRef: nodeRef},
+	}
 	newest := &clusterv1.Machine{
 		ObjectMeta: metav1.ObjectMeta{CreationTimestamp: metav1.NewTime(currentTime.Time.AddDate(0, 0, -1))},
 		Status:     clusterv1.MachineStatus{NodeRef: nodeRef},
@@ -513,6 +517,14 @@ func TestMachineOldestDelete(t *testing.T) {
 			},
 			expect: []*clusterv1.Machine{nodeHealthyConditionUnknownMachine},
 		},
+		{
+			desc: "func=oldestDeletePriority, diff=1 (deletionTimestamp)",
+			diff: 1,
+			machines: []*clusterv1.Machine{
+				empty, secondNewest, oldest, secondOldest, newest, nodeHealthyConditionUnknownMachine, mustDeleteMachine,
+			},
+			expect: []*clusterv1.Machine{mustDeleteMachine},
+		},
 	}
 
 	for _, test := range tests {