Add reconciler to shutdown a subcluster

.github/actions/run-e2e-leg/action.yaml

@@ -60,6 +60,10 @@ inputs:
     description: 'Whether BASE_VERTICA_IMG is required'
     required: false
     default: 'false'
+  vertica-license:
+    description: 'Contents of a vertica license to use with the tests'
+    required: false
+    default: ''
 runs:
   using: "composite"
   steps:
@@ -94,6 +98,11 @@ runs:
         export OPERATOR_IMG=${{ inputs.operator-image }}
         export VLOGGER_IMG=${{ inputs.vlogger-image }}
 
+        if [[ "${{ inputs.vertica-license }}" != "" ]]; then
+          export LICENSE_FILE=/tmp/vertica-license.dat
+          echo -n "${{ inputs.vertica-license }}" > $LICENSE_FILE
+        fi
+
         # Set DEPLOY_WITH
         export DEPLOY_WITH="${{ inputs.deploy-with }}";
 

.github/workflows/e2e.yml

@@ -50,6 +50,7 @@ on:
         - vcluster leg 6
         - vcluster leg 7
         - leg 8
+        - vcluster leg 9
         - vcluster server upgrade
         - vcluster udx
       run_security_scan:
@@ -436,6 +437,29 @@ jobs:
           communal-storage-type: azb
           minimum-vertica-image: '24.2.0'
 
+  e2e-leg-9-vcluster:
+    if: ${{ ! github.event.pull_request.head.repo.fork  && (inputs.e2e_test_suites == 'all' || inputs.e2e_test_suites == 'vcluster leg 9' || inputs.e2e_test_suites == '')}}
+    needs: [build]
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Test
+        uses: ./.github/actions/run-e2e-leg
+        with:
+          leg-identifier: 'leg-9'
+          dockerhub-username: ${{ secrets.DOCKERHUB_USERNAME }}
+          dockerhub-token: ${{ secrets.DOCKERHUB_TOKEN }}
+          vlogger-image: ${{ needs.build.outputs.vlogger-image }}
+          operator-image: ${{ needs.build.outputs.operator-image }}
+          vertica-image: ${{ needs.build.outputs.minimal-vertica-image }}
+          vertica-deployment-method: vclusterops
+          communal-storage-type: s3
+          minimum-vertica-image: '24.3.0'
+          # Include the vertica license so we can test scaling past 3 nodes.
+          vertica-license: ${{ secrets.VERTICA_LICENSE }}
+          need-base-vertica-image: 'true'
+
   e2e-server-upgrade-admintools:
     if: ${{ inputs.e2e_test_suites == 'all' || inputs.e2e_test_suites == 'admintools server upgrade' || inputs.e2e_test_suites == '' }}
     needs: [build]
@@ -536,4 +560,4 @@ jobs:
           # We must pass the full image since the minimal image cannot run Java UDx's
           vertica-image: ${{ needs.build.outputs.full-vertica-image }}
           vertica-deployment-method: vclusterops
-          communal-storage-type: s3
+          communal-storage-type: s3

Makefile

@@ -213,11 +213,15 @@ CONCURRENCY_VERTICAAUTOSCALER?=1
 CONCURRENCY_EVENTTRIGGER?=1
 CONCURRENCY_VERTICARESTOREPOINTSQUERY?=1
 CONCURRENCY_VERTICASCRUTINIZE?=1
+CONCURRENCY_SANDBOXCONFIGMAP?=1
+CONCURRENCY_VERTICAREPLICATOR?=1
 export CONCURRENCY_VERTICADB \
   CONCURRENCY_VERTICAAUTOSCALER \
   CONCURRENCY_EVENTTRIGGER \
   CONCURRENCY_VERTICARESTOREPOINTSQUERY \
-  CONCURRENCY_VERTICASCRUTINIZE
+  CONCURRENCY_VERTICASCRUTINIZE \
+  CONCURRENCY_SANDBOXCONFIGMAP \
+  CONCURRENCY_VERTICAREPLICATOR
 
 # Clear this variable if you don't want to wait for the helm deployment to
 # finish before returning control. This exists to allow tests to attempt deploy

PROJECT

@@ -71,4 +71,12 @@ resources:
     defaulting: true
     validation: true
     webhookVersion: v1
+- api:
+    crdVersion: v1
+    namespaced: true
+  controller: true
+  domain: vertica.com
+  kind: VerticaReplicator
+  path: github.com/vertica/vertica-kubernetes/api/v1beta1
+  version: v1beta1
 version: "3"

api/v1/helpers.go

@@ -42,6 +42,8 @@ const (
 
 	RFC1123DNSSubdomainNameRegex = `^[a-z0-9]([-a-z0-9]*[a-z0-9])?(\.[a-z0-9]([-a-z0-9]*[a-z0-9])?)*$`
 	RFC1035DNSLabelNameRegex     = `^[a-z]([a-z0-9\-]{0,61}[a-z0-9])?$`
+
+	MainCluster = ""
 )
 
 // ExtractNamespacedName gets the name and returns it as a NamespacedName
@@ -57,6 +59,20 @@ func MakeVDBName() types.NamespacedName {
 	return types.NamespacedName{Name: "vertica-sample", Namespace: "default"}
 }
 
+// GenerateOwnerReference creates an owner reference for the current VerticaDB
+func (v *VerticaDB) GenerateOwnerReference() metav1.OwnerReference {
+	isController := true
+	blockOwnerDeletion := false
+	return metav1.OwnerReference{
+		APIVersion:         GroupVersion.String(),
+		Kind:               VerticaDBKind,
+		Name:               v.Name,
+		UID:                v.GetUID(),
+		Controller:         &isController,
+		BlockOwnerDeletion: &blockOwnerDeletion,
+	}
+}
+
 // FindTransientSubcluster will return a pointer to the transient subcluster if one exists
 func (v *VerticaDB) FindTransientSubcluster() *Subcluster {
 	for i := range v.Spec.Subclusters {
@@ -125,11 +141,19 @@ func MakeVDBForHTTP(httpServerTLSSecretName string) *VerticaDB {
 // vclusterops. This is intended for test purposes.
 func MakeVDBForVclusterOps() *VerticaDB {
 	vdb := MakeVDB()
-	vdb.Annotations[vmeta.VersionAnnotation] = ScrutinizeDBPasswdInSecretMinVersion
+	vdb.Annotations[vmeta.VersionAnnotation] = VcluseropsAsDefaultDeploymentMethodMinVersion
 	vdb.Annotations[vmeta.VClusterOpsAnnotation] = vmeta.VClusterOpsAnnotationTrue
 	return vdb
 }
 
+// MakeVDBForScrutinize is a helper that constructs a VerticaDB struct for
+// scrutinize. This is intended for test purposes.
+func MakeVDBForScrutinize() *VerticaDB {
+	vdb := MakeVDBForVclusterOps()
+	vdb.Annotations[vmeta.VersionAnnotation] = ScrutinizeDBPasswdInSecretMinVersion
+	return vdb
+}
+
 // GenSubclusterMap will organize all of the subclusters into a map for quicker lookup
 func (v *VerticaDB) GenSubclusterMap() map[string]*Subcluster {
 	scMap := map[string]*Subcluster{}
@@ -368,7 +392,7 @@ func (v *VerticaDB) IsKnownDepotVolumeType() bool {
 	return false
 }
 
-// getFirstPrimarySubcluster returns the first primary subcluster defined in the vdb
+// GetFirstPrimarySubcluster returns the first primary subcluster defined in the vdb
 func (v *VerticaDB) GetFirstPrimarySubcluster() *Subcluster {
 	for i := range v.Spec.Subclusters {
 		sc := &v.Spec.Subclusters[i]
@@ -380,6 +404,68 @@ func (v *VerticaDB) GetFirstPrimarySubcluster() *Subcluster {
 	return nil
 }
 
+// HasSecondarySubclusters returns true if at least 1 secondary subcluster
+// exists in the database.
+func (v *VerticaDB) HasSecondarySubclusters() bool {
+	for i := range v.Spec.Subclusters {
+		if v.Spec.Subclusters[i].IsSecondary() {
+			return true
+		}
+	}
+	return false
+}
+
+// IsAutoUpgradePolicy returns true
+func (v *VerticaDB) IsAutoUpgradePolicy() bool {
+	return v.Spec.UpgradePolicy == "" || v.Spec.UpgradePolicy == AutoUpgrade
+}
+
+// GetUpgradePolicyToUse returns the upgrade policy that the db should use.
+// It will take into account the settings in the vdb as well as what is
+// supported in the server. This will never return the auto upgrade policy. If
+// you need the current value of that field, just refer to it by referencing
+// Spec.UpgradePolicy.
+func (v *VerticaDB) GetUpgradePolicyToUse() UpgradePolicyType {
+	if v.Spec.UpgradePolicy == OfflineUpgrade {
+		return OfflineUpgrade
+	}
+
+	if v.IsAutoUpgradePolicy() && v.IsKSafety0() {
+		return OfflineUpgrade
+	}
+
+	// If we cannot get the version, always default to offline. We cannot make
+	// any assumptions about what upgrade policy the server supports.
+	vinf, ok := v.MakeVersionInfo()
+	if !ok {
+		return OfflineUpgrade
+	}
+
+	// The Replicated option can only be chosen explicitly. Although eventually,
+	// the Auto option will automatically select this method, we first need to
+	// complete the implementation of this new policy.
+	if v.Spec.UpgradePolicy == ReplicatedUpgrade {
+		// Replicated upgrade requires that we scale out the cluster. See if
+		// there is evidence that we have already scaled past 3 nodes (CE
+		// license limit), or we have a license defined.
+		const ceLicenseLimit = 3
+		if vinf.IsEqualOrNewer(ReplicatedUpgradeVersion) &&
+			!v.IsKSafety0() &&
+			(v.getNumberOfNodes() > ceLicenseLimit || v.Spec.LicenseSecret != "") {
+			return ReplicatedUpgrade
+		} else if vinf.IsEqualOrNewer(OnlineUpgradeVersion) {
+			return OnlineUpgrade
+		}
+	}
+
+	if (v.Spec.UpgradePolicy == OnlineUpgrade || v.IsAutoUpgradePolicy()) &&
+		vinf.IsEqualOrNewer(OnlineUpgradeVersion) {
+		return OnlineUpgrade
+	}
+
+	return OfflineUpgrade
+}
+
 // GetIgnoreClusterLease will check if the cluster lease should be ignored.
 func (v *VerticaDB) GetIgnoreClusterLease() bool {
 	return vmeta.IgnoreClusterLease(v.Annotations)
@@ -662,3 +748,36 @@ func (v *VerticaDB) IsHTTPSTLSConfGenerationEnabled() (bool, error) {
 	}
 	return !inf.IsEqualOrNewer(AutoGenerateHTTPSCertsForNewDatabasesMinVersion), nil
 }
+
+// GetSubclusterSandboxName returns the sandbox the given subcluster belongs to,
+// or an empty string if it does not belong to any
+func (v *VerticaDB) GetSubclusterSandboxName(scName string) string {
+	for i := range v.Spec.Sandboxes {
+		for j := range v.Spec.Sandboxes[i].Subclusters {
+			if scName == v.Spec.Sandboxes[i].Subclusters[j].Name {
+				return v.Spec.Sandboxes[i].Name
+			}
+		}
+	}
+	return MainCluster
+}
+
+// getNumberOfNodes returns the number of nodes defined in the database, as per the CR.
+func (v *VerticaDB) getNumberOfNodes() int {
+	count := 0
+	for i := range v.Spec.Subclusters {
+		count += int(v.Spec.Subclusters[i].Size)
+	}
+	return count
+}
+
+// GetSandbox returns the sandbox given by name. A nil pointer is returned if
+// not found.
+func (v *VerticaDB) GetSandbox(sbName string) *Sandbox {
+	for i := range v.Spec.Sandboxes {
+		if v.Spec.Sandboxes[i].Name == sbName {
+			return &v.Spec.Sandboxes[i]
+		}
+	}
+	return nil
+}

api/v1/version.go

@@ -32,6 +32,8 @@ const (
 	NodesHaveReadOnlyStateVersion = "v11.0.2"
 	// The minimum version that allows for online upgrade.
 	OnlineUpgradeVersion = "v11.1.0"
+	// The minimum version that allows for replicated upgrade.
+	ReplicatedUpgradeVersion = "v24.3.0"
 	// The version that added the --force option to reip to handle up nodes
 	ReIPAllowedWithUpNodesVersion = "v11.1.0"
 	// The version of the server that doesn't support cgroup v2
@@ -43,6 +45,8 @@ const (
 	// The minimum version that we can use the option with create DB to skip the
 	// package install.
 	CreateDBSkipPackageInstallVersion = "v12.0.1"
+	// The version that added sandbox state
+	NodesHaveSandboxStateVersion = "v12.0.4"
 	// Starting in v23.4.0, we added some new config parameters for settings
 	// that were typically done post create using SQL -- setting the default
 	// subcluster name and preferred k-safety.
@@ -62,6 +66,8 @@ const (
 	InDatabaseLogRotateMinVersion = "v24.1.0"
 	// Starting in v24.2.0, restoring from a restore point in archive is supported.
 	RestoreSupportedMinVersion = "v24.2.0"
+	// Starting in v24.3.0, database replication via vclusterops is supported.
+	ReplicationViaVclusteropsSupportedMinVersion = "v24.3.0"
 	// The NMA TLS secret can be stored in an external secret store. These are
 	// the minimum versions of the NMA that we support them.
 	NMATLSSecretInGSMMinVersion               = "v24.1.0"
@@ -71,6 +77,10 @@ const (
 	// Starting in v24.2.0, vcluster scrutinize command can read the
 	// database password from secret(k8s, aws, gsm)
 	ScrutinizeDBPasswdInSecretMinVersion = "v24.2.0"
+	// Starting in v24.3.0, sandboxing a subcluster with the operator is supported
+	SandboxSupportedMinVersion = "v24.3.0"
+	// Starting in v24.3.0, we call vclusterops API to get node details instead of executing vsql within the pod
+	FetchNodeDetailsWithVclusterOpsMinVersion = "v24.3.0"
 )
 
 // GetVerticaVersionStr returns the vertica version, in string form, that is stored