controller: Add migration label to pause reconciliation

To support migrating CockroachDB clusters from the public operator to the
CockroachDB operator through coexistence, we need to prevent the operator
from reverting manual changes made during the migration process. This includes
preventing the scale-up of StatefulSet pods after manual scale-down and
allowing modifications to Service labels (e.g., public service) which would
otherwise be reset by the operator.

This commit introduces a new migration label (`crdb.io/migrating`) in the
v1alpha1 API. When this label is set to "true" on the CrdbCluster resource,
the controller stops the entire reconciliation loop. This ensures the
operator does not interfere with any resources during the migration phase.

Author: NishanthNalluri

Makefile

@@ -373,4 +373,3 @@ publish-operator-openshift:
 .PHONY: release/publish-openshift-bundle
 release/publish-openshift-bundle:
 	./build/release/teamcity-publish-openshift-bundle.sh
-

apis/v1alpha1/cluster_types.go

@@ -22,6 +22,13 @@ import (
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 )
 
+const (
+	// CrdbOperatorMigrationLabel is the label used to indicate that the cluster is migrating.
+	// If this label is set to "true", the operator will stop reconciling the
+	// entire cluster.
+	CrdbOperatorMigrationLabel = "crdb.io/skip-reconcile"
+)
+
 // NOTE: json tags are required.  Any new fields you add must have json tags for the fields to be serialized.
 // Important: Run "make dev/generate" to regenerate code after modifying this file
 

pkg/controller/cluster_controller.go

@@ -142,6 +142,12 @@ func (r *ClusterReconciler) Reconcile(ctx context.Context, req reconcile.Request
 		}
 	}
 
+	// If the cluster is migrating, we stop reconciliation
+	if val, ok := cluster.Unwrap().Labels[api.CrdbOperatorMigrationLabel]; ok && val == "true" {
+		log.Info("cluster is migrating, stopping reconciliation")
+		return noRequeue()
+	}
+
 	actorToExecute, err := r.Director.GetActorToExecute(ctx, &cluster, log)
 	if err != nil {
 		return requeueAfter(30*time.Second, nil)

pkg/controller/cluster_controller_test.go

@@ -147,5 +147,44 @@ func TestReconcile(t *testing.T) {
 			assert.Equal(t, tt.want, actual)
 		})
 	}
+}
+
+func TestReconcileMigrationLabel(t *testing.T) {
+	scheme := testutil.InitScheme(t)
+
+	ns := &corev1.Namespace{
+		ObjectMeta: metav1.ObjectMeta{
+			Name: "test-namespace",
+		},
+	}
+
+	cluster := testutil.NewBuilder("cluster").Namespaced(ns.Name).WithNodeCount(1).Cr()
+	// Set status so we skip the "first reconcile" block
+	cluster.Status.ClusterStatus = "Running"
+	// Set the migration label
+	cluster.Labels = map[string]string{
+		api.CrdbOperatorMigrationLabel: "true",
+	}
+
+	objs := []runtime.Object{
+		ns,
+		cluster,
+	}
+
+	cl := fake.NewClientBuilder().WithScheme(scheme).WithRuntimeObjects(objs...).WithStatusSubresource(cluster).Build()
+	log := zapr.NewLogger(zaptest.NewLogger(t)).WithName("cluster-controller-test")
+	req := ctrl.Request{NamespacedName: types.NamespacedName{Namespace: cluster.Namespace, Name: cluster.Name}}
+
+	r := &controller.ClusterReconciler{
+		Client: cl,
+		Log:    log,
+		Scheme: scheme,
+		Director: &fakeDirector{
+			actorToExecute: &fakeActor{},
+		},
+	}
 
+	actual, err := r.Reconcile(context.TODO(), req)
+	require.NoError(t, err)
+	assert.Equal(t, ctrl.Result{}, actual)
 }

pkg/resource/statefulset_test.go

@@ -22,6 +22,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"testing"
 
 	api "github.com/cockroachdb/cockroach-operator/apis/v1alpha1"
 	"github.com/cockroachdb/cockroach-operator/pkg/labels"
@@ -32,8 +33,6 @@ import (
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 	appsv1 "k8s.io/api/apps/v1"
-
-	"testing"
 )
 
 var update = flag.Bool("update", false, "update the golden files of this test")