Add e2e test for changing a cluster to use DNS in cluster files

controllers/change_coordinators.go

@@ -108,7 +108,6 @@ func (c changeCoordinators) reconcile(ctx context.Context, r *FoundationDBCluste
 	return nil
 }
 
-// TODO move them into separate package?
 // selectCandidates is a helper for Reconcile that picks non-excluded, not-being-removed class-matching process groups.
 func selectCandidates(cluster *fdbv1beta2.FoundationDBCluster, status *fdbv1beta2.FoundationDBStatus) ([]locality.Info, error) {
 	candidates := make([]locality.Info, 0, len(status.Cluster.Processes))
@@ -121,6 +120,19 @@ func selectCandidates(cluster *fdbv1beta2.FoundationDBCluster, status *fdbv1beta
 			continue
 		}
 
+		// Ignore processes with missing locality, see: https://github.com/FoundationDB/fdb-kubernetes-operator/issues/1254
+		if len(process.Locality) == 0 {
+			continue
+		}
+
+		// If the cluster should be using DNS in the cluster file we should make sure the locality is set.
+		if cluster.UseDNSInClusterFile() {
+			_, ok := process.Locality[fdbv1beta2.FDBLocalityDNSNameKey]
+			if !ok {
+				continue
+			}
+		}
+
 		if cluster.ProcessGroupIsBeingRemoved(fdbv1beta2.ProcessGroupID(process.Locality[fdbv1beta2.FDBLocalityInstanceIDKey])) {
 			continue
 		}
@@ -187,5 +199,6 @@ func getCoordinatorAddress(cluster *fdbv1beta2.FoundationDBCluster, locality loc
 			Flags:         address.Flags,
 		}
 	}
+
 	return address
 }

e2e/test_operator/operator_test.go

@@ -1353,6 +1353,35 @@ var _ = Describe("Operator", Label("e2e", "pr"), func() {
 		})
 	})
 
+	When("migrating a cluster to make use of DNS in the cluster file", func() {
+		BeforeEach(func() {
+			cluster := fdbCluster.GetCluster()
+			parsedVersion, err := fdbv1beta2.ParseFdbVersion(cluster.Status.RunningVersion)
+			Expect(err).NotTo(HaveOccurred())
+
+			if !parsedVersion.SupportsDNSInClusterFile() {
+				Skip(fmt.Sprintf("current FoundationDB version %s doesn't support DNS", parsedVersion.String()))
+			}
+
+			if cluster.UseDNSInClusterFile() {
+				Skip("cluster already uses DNS")
+			}
+
+			Expect(fdbCluster.SetUseDNSInClusterFile(true)).ToNot(HaveOccurred())
+		})
+
+		It("should migrate the cluster", func() {
+			cluster := fdbCluster.GetCluster()
+			Eventually(func() string {
+				return fdbCluster.GetStatus().Cluster.ConnectionString
+			}).Should(ContainSubstring(cluster.GetDNSDomain()))
+		})
+
+		AfterEach(func() {
+			Expect(fdbCluster.SetUseDNSInClusterFile(false)).ToNot(HaveOccurred())
+		})
+	})
+
 	// This test is pending, as all Pods will be restarted at the same time, which will lead to unavailability without
 	// using DNS.
 	PWhen("crash looping the sidecar for all Pods", func() {