fix: skip not found nodes when reconciling LB backend address pools

Author: nilo19

staging/src/k8s.io/legacy-cloud-providers/azure/azure_vmss.go

@@ -964,6 +964,11 @@ func (ss *scaleSet) EnsureHostInPool(service *v1.Service, nodeName types.NodeNam
 	vmName := mapNodeNameToVMName(nodeName)
 	ssName, instanceID, vm, err := ss.getVmssVM(vmName, azcache.CacheReadTypeDefault)
 	if err != nil {
+		if errors.Is(err, cloudprovider.InstanceNotFound) {
+			klog.Infof("EnsureHostInPool: skipping node %s because it is not found", vmName)
+			return "", "", "", nil, nil
+		}
+
 		return "", "", "", nil, err
 	}
 
@@ -1339,6 +1344,11 @@ func (ss *scaleSet) EnsureHostsInPool(service *v1.Service, nodes []*v1.Node, bac
 func (ss *scaleSet) ensureBackendPoolDeletedFromNode(nodeName, backendPoolID string) (string, string, string, *compute.VirtualMachineScaleSetVM, error) {
 	ssName, instanceID, vm, err := ss.getVmssVM(nodeName, azcache.CacheReadTypeDefault)
 	if err != nil {
+		if errors.Is(err, cloudprovider.InstanceNotFound) {
+			klog.Infof("ensureBackendPoolDeletedFromNode: skipping node %s because it is not found", nodeName)
+			return "", "", "", nil, nil
+		}
+
 		return "", "", "", nil, err
 	}
 

staging/src/k8s.io/legacy-cloud-providers/azure/azure_vmss_test.go

@@ -2144,7 +2144,7 @@ func TestEnsureHostsInPool(t *testing.T) {
 			expectedVMSSVMPutTimes: 1,
 		},
 		{
-			description: "EnsureHostsInPool should gather report the error if something goes wrong in EnsureHostInPool",
+			description: "EnsureHostsInPool should skip not found nodes",
 			nodes: []*v1.Node{
 				{
 					ObjectMeta: metav1.ObjectMeta{
@@ -2158,7 +2158,7 @@ func TestEnsureHostsInPool(t *testing.T) {
 			backendpoolID:          testLBBackendpoolID1,
 			vmSetName:              testVMSSName,
 			expectedVMSSVMPutTimes: 0,
-			expectedErr:            true,
+			expectedErr:            false,
 		},
 	}
 
@@ -2204,9 +2204,8 @@ func TestEnsureBackendPoolDeletedFromNode(t *testing.T) {
 		expectedErr               error
 	}{
 		{
-			description: "ensureBackendPoolDeletedFromNode should report the error that occurs during getVmssVM",
+			description: "ensureBackendPoolDeletedFromNode should skip not found nodes",
 			nodeName:    "vmss-vm-000001",
-			expectedErr: cloudprovider.InstanceNotFound,
 		},
 		{
 			description:           "ensureBackendPoolDeletedFromNode skip the node if the VM's NIC config is nil",