k8s: Put brokers in maintenance mode before deleting orphan pod

During rolling update, before this change, Redpanda operator was calculating
the difference between running pod specification and stateful set pod template.
If the specification did not match the pod was deleted. From release v22.1.1
operator is configuring each broker with pod lifecycle hooks. In the PreStop
hook the script will try to put broker into maintenance mode for 120 seconds
before POD is terminated. Redpanda could not finish within 120 seconds to put
one broker into maintenance mode.

This PR improves the situation by putting maintenance mode before POD is
deleted. The `EnableMaintenanceMode` function is called multiple times until
`Broker` function returns correct status. The assumption is that REST admin API
maintenance mode endpoint is idempotent.

When pod is successfully deleted statefulset would reschedule the pod with
correct pod specification.

redpanda-data#4125
redpanda-data#3023

src/go/k8s/pkg/admin/admin.go

@@ -92,6 +92,7 @@ type AdminAPIClient interface {
 	GetLicenseInfo(ctx context.Context) (admin.License, error)
 
 	Brokers(ctx context.Context) ([]admin.Broker, error)
+	Broker(ctx context.Context, nodeID int) (admin.Broker, error)
 	DecommissionBroker(ctx context.Context, node int) error
 	RecommissionBroker(ctx context.Context, node int) error
 

src/go/k8s/pkg/admin/mock_admin.go

@@ -390,6 +390,26 @@ func (m *MockAdminAPI) SetBrokerStatus(
 	return fmt.Errorf("unknown broker %d", id)
 }
 
+func (m *MockAdminAPI) Broker(_ context.Context, nodeID int) (admin.Broker, error) {
+	t := true
+	return admin.Broker{
+		NodeID:           nodeID,
+		NumCores:         2,
+		MembershipStatus: "",
+		IsAlive:          &t,
+		Version:          "unversioned",
+		Maintenance: &admin.MaintenanceStatus{
+			Draining:     false,
+			Finished:     true,
+			Errors:       false,
+			Partitions:   0,
+			Eligible:     0,
+			Transferring: 0,
+			Failed:       0,
+		},
+	}, nil
+}
+
 func makeCopy(input, output interface{}) {
 	ser, err := json.Marshal(input)
 	if err != nil {

src/go/k8s/pkg/resources/statefulset_update.go

@@ -18,6 +18,7 @@ import (
 	"net/url"
 	"reflect"
 	"sort"
+	"strconv"
 	"strings"
 	"time"
 
@@ -173,6 +174,22 @@ func (r *StatefulSetResource) rollingUpdate(
 		}
 
 		if !patchResult.IsEmpty() {
+			var ordinal int64
+			ordinal, err = strconv.ParseInt(pod.Name[len(r.pandaCluster.Name)+1:], 10, 32)
+			if err != nil {
+				return fmt.Errorf("cluster %s: cannot convert pod name (%s) to ordinal: %w", r.pandaCluster.Name, pod.Name, err)
+			}
+
+			if err = r.putInMaintenanceMode(ctx, int32(ordinal)); err != nil {
+				// As maintenance mode can not be easily watched using controller runtime the requeue error
+				// is always returned. That way a rolling update will not finish when operator waits for
+				// maintenance mode finished.
+				return &RequeueAfterError{
+					RequeueAfter: RequeueDuration,
+					Msg:          fmt.Sprintf("putting node (%s) into maintenance mode: %v", pod.Name, err),
+				}
+			}
+
 			r.logger.Info("Changes in Pod definition other than activeDeadlineSeconds, configurator and Redpanda container name. Deleting pod",
 				"pod-name", pod.Name,
 				"patch", patchResult.Patch)
@@ -207,6 +224,38 @@ func (r *StatefulSetResource) rollingUpdate(
 	return nil
 }
 
+//nolint:goerr113 // out of scope for this PR
+func (r *StatefulSetResource) putInMaintenanceMode(ctx context.Context, ordinal int32) error {
+	adminAPIClient, err := r.getAdminAPIClient(ctx, ordinal)
+	if err != nil {
+		return fmt.Errorf("creating admin API client: %w", err)
+	}
+
+	nodeConf, err := adminAPIClient.GetNodeConfig(ctx)
+	if err != nil {
+		return fmt.Errorf("getting node config: %w", err)
+	}
+
+	err = adminAPIClient.EnableMaintenanceMode(ctx, nodeConf.NodeID)
+	if err != nil {
+		return fmt.Errorf("enabling maintenance mode: %w", err)
+	}
+
+	br, err := adminAPIClient.Broker(ctx, nodeConf.NodeID)
+	if err != nil {
+		return fmt.Errorf("getting broker infromations: %w", err)
+	}
+	if br.Maintenance != nil &&
+		br.Maintenance.Finished &&
+		!br.Maintenance.Draining &&
+		!br.Maintenance.Errors &&
+		br.Maintenance.Failed == 0 {
+		return fmt.Errorf("maintenance mode is not valid to do rolling update: %v", br.Maintenance)
+	}
+
+	return nil
+}
+
 func (r *StatefulSetResource) updateStatefulSet(
 	ctx context.Context,
 	current *appsv1.StatefulSet,