feat: Update operator to support Hyperbytedb 0.8.3

api/v1alpha1/hyperbytedbcluster_types.go

@@ -29,9 +29,15 @@ type HyperbytedbClusterSpec struct {
 	// +kubebuilder:default=1
 	Replicas *int32 `json:"replicas,omitempty"`
 
-	// +kubebuilder:default="hyperbytedb:latest"
+	// Container image reference. When set with a tag (contains ':'), used as-is.
+	// When set without a tag, treated as the repository and combined with Version.
+	// When empty, defaults to hyperbytedb:{Version} or hyperbytedb:latest.
+	// +optional
 	Image string `json:"image,omitempty"`
 
+	// Application version. Drives the container image tag for hyperbytedb and
+	// hyperbytedb-proxy (e.g. version "0.8.3" → hyperbytedb:v0.8.3). Changing
+	// this field triggers a rolling upgrade.
 	// +optional
 	Version string `json:"version,omitempty"`
 
@@ -169,6 +175,12 @@ type CertManagerIssuerRef struct {
 type StorageSpec struct {
 	// +optional
 	VolumeClaimTemplate *PersistentVolumeClaimSpec `json:"volumeClaimTemplate,omitempty"`
+
+	// WAL encoding format: "bincode" (default) or "arrow_ipc".
+	// +optional
+	// +kubebuilder:validation:Enum=bincode;arrow_ipc
+	// +kubebuilder:default="bincode"
+	WALFormat string `json:"walFormat,omitempty"`
 }
 
 type PersistentVolumeClaimSpec struct {
@@ -204,6 +216,11 @@ type FlushSpec struct {
 	// +optional
 	// +kubebuilder:validation:Minimum=0
 	WALBatchDelayUs int64 `json:"walBatchDelayUs,omitempty"`
+
+	// Keep chDB-ready Arrow batches in an in-memory WAL cache for zero-copy flush.
+	// +optional
+	// +kubebuilder:default=true
+	ArrowWALEnabled *bool `json:"arrowWALEnabled,omitempty"`
 }
 
 type ChDBSpec struct {
@@ -235,18 +252,6 @@ type LoggingSpec struct {
 	// +kubebuilder:default="text"
 	// +kubebuilder:validation:Enum=text;json
 	Format string `json:"format,omitempty"`
-
-	// Emit per-phase performance logs (write/query/flush). Off by default.
-	// +optional
-	DetailedTrace *bool `json:"detailedTrace,omitempty"`
-
-	// OTLP HTTP endpoint for trace export (e.g. http://alloy-logs:4318).
-	// +optional
-	OtlpEndpoint string `json:"otlpEndpoint,omitempty"`
-
-	// Fraction of traces exported to OTLP (0.0–1.0). Default 1.0 when OTLP is set.
-	// +optional
-	OtlpSampleRatio string `json:"otlpSampleRatio,omitempty"`
 }
 
 type ClusterTuningSpec struct {
@@ -447,7 +452,7 @@ type ProxySpec struct {
 	// When false, the operator does not create or reconcile any proxy
 	// resources. Existing proxy Deployment/Service (if any) are left alone
 	// so they can be cleaned up out-of-band.
-	// +kubebuilder:default=false
+	// +kubebuilder:default=true
 	Enabled bool `json:"enabled"`
 
 	// +kubebuilder:default="hyperbytedb-proxy:latest"
@@ -479,7 +484,7 @@ type ProxySpec struct {
 	// request with 503. Bigger values mean rolling restarts are smoother but
 	// individual stuck requests sit longer.
 	// +optional
-	// +kubebuilder:default=10
+	// +kubebuilder:default=30
 	// +kubebuilder:validation:Minimum=0
 	HoldTimeoutSecs int32 `json:"holdTimeoutSecs,omitempty"`
 

api/v1alpha1/hyperbytedbcluster_webhook.go

@@ -43,10 +43,6 @@ func (w *HyperbytedbClusterWebhook) Default(_ context.Context, obj *HyperbytedbC
 		obj.Spec.Replicas = ptr.To(int32(1))
 	}
 
-	if obj.Spec.Image == "" {
-		obj.Spec.Image = "hyperbytedb:latest"
-	}
-
 	if obj.Spec.Server.Port == 0 {
 		obj.Spec.Server.Port = 8086
 	}
@@ -84,6 +80,12 @@ func (w *HyperbytedbClusterWebhook) Default(_ context.Context, obj *HyperbytedbC
 		obj.Spec.Retention.Interval = "60s"
 	}
 
+	// hyperbytedb-proxy is enabled by default for health-aware routing during
+	// rolling upgrades. Opt out with spec.proxy.enabled=false.
+	if obj.Spec.Proxy == nil {
+		obj.Spec.Proxy = &ProxySpec{Enabled: true}
+	}
+
 	return nil
 }
 

config/crd/bases/hyperbytedb.hyperbytedb.io_hyperbytedbclusters.yaml

@@ -3137,6 +3137,11 @@ spec:
                 type: object
               flush:
                 properties:
+                  arrowWALEnabled:
+                    default: true
+                    description: Keep chDB-ready Arrow batches in an in-memory WAL
+                      cache for zero-copy flush.
+                    type: boolean
                   intervalSecs:
                     default: 10
                     format: int32
@@ -3189,7 +3194,10 @@ spec:
                     type: integer
                 type: object
               image:
-                default: hyperbytedb:latest
+                description: |-
+                  Container image reference. When set with a tag (contains ':'), used as-is.
+                  When set without a tag, treated as the repository and combined with Version.
+                  When empty, defaults to hyperbytedb:{Version} or hyperbytedb:latest.
                 type: string
               imagePullPolicy:
                 description: PullPolicy describes a policy for if/when to pull a container
@@ -3215,10 +3223,6 @@ spec:
                 type: array
               logging:
                 properties:
-                  detailedTrace:
-                    description: Emit per-phase performance logs (write/query/flush).
-                      Off by default.
-                    type: boolean
                   format:
                     default: text
                     enum:
@@ -3234,13 +3238,6 @@ spec:
                     - warn
                     - error
                     type: string
-                  otlpEndpoint:
-                    description: OTLP HTTP endpoint for trace export (e.g. http://alloy-logs:4318).
-                    type: string
-                  otlpSampleRatio:
-                    description: Fraction of traces exported to OTLP (0.0–1.0). Default
-                      1.0 when OTLP is set.
-                    type: string
                 type: object
               monitoring:
                 properties:
@@ -3271,7 +3268,7 @@ spec:
                   returning errors to clients.
                 properties:
                   enabled:
-                    default: false
+                    default: true
                     description: |-
                       When false, the operator does not create or reconcile any proxy
                       resources. Existing proxy Deployment/Service (if any) are left alone
@@ -3283,7 +3280,7 @@ spec:
                       Set to `/health/ready` for the deeper chDB-aware readiness check.
                     type: string
                   holdTimeoutSecs:
-                    default: 10
+                    default: 30
                     description: |-
                       How long the proxy waits for a backend to come back before failing a
                       request with 503. Bigger values mean rolling restarts are smoother but
@@ -3609,6 +3606,13 @@ spec:
                       storageClassName:
                         type: string
                     type: object
+                  walFormat:
+                    default: bincode
+                    description: 'WAL encoding format: "bincode" (default) or "arrow_ipc".'
+                    enum:
+                    - bincode
+                    - arrow_ipc
+                    type: string
                 type: object
               tolerations:
                 items:
@@ -3823,6 +3827,10 @@ spec:
                   type: object
                 type: array
               version:
+                description: |-
+                  Application version. Drives the container image tag for hyperbytedb and
+                  hyperbytedb-proxy (e.g. version "0.8.3" → hyperbytedb:v0.8.3). Changing
+                  this field triggers a rolling upgrade.
                 type: string
             type: object
           status:

config/samples/hyperbytedb_v1alpha1_hyperbytedbcluster_ha.yaml

@@ -27,8 +27,6 @@ spec:
   logging:
     level: info
     format: json
-    otlpEndpoint: http://alloy-logs:4318
-    otlpSampleRatio: "0.1"
   cardinality:
     maxTagValuesPerMeasurement: 100000
     maxMeasurementsPerDatabase: 10000

dist/chart/templates/crd/hyperbytedbclusters.hyperbytedb.hyperbytedb.io.yaml

@@ -1,4 +1,5 @@
 ---
+---
 apiVersion: apiextensions.k8s.io/v1
 kind: CustomResourceDefinition
 metadata:
@@ -3189,7 +3190,10 @@ spec:
                     type: integer
                 type: object
               image:
-                default: hyperbytedb:latest
+                description: |-
+                  Container image reference. When set with a tag (contains ':'), used as-is.
+                  When set without a tag, treated as the repository and combined with Version.
+                  When empty, defaults to hyperbytedb:{Version} or hyperbytedb:latest.
                 type: string
               imagePullPolicy:
                 description: PullPolicy describes a policy for if/when to pull a container
@@ -3271,7 +3275,7 @@ spec:
                   returning errors to clients.
                 properties:
                   enabled:
-                    default: false
+                    default: true
                     description: |-
                       When false, the operator does not create or reconcile any proxy
                       resources. Existing proxy Deployment/Service (if any) are left alone
@@ -3283,7 +3287,7 @@ spec:
                       Set to `/health/ready` for the deeper chDB-aware readiness check.
                     type: string
                   holdTimeoutSecs:
-                    default: 10
+                    default: 30
                     description: |-
                       How long the proxy waits for a backend to come back before failing a
                       request with 503. Bigger values mean rolling restarts are smoother but
@@ -3823,6 +3827,10 @@ spec:
                   type: object
                 type: array
               version:
+                description: |-
+                  Application version. Drives the container image tag for hyperbytedb and
+                  hyperbytedb-proxy (e.g. version "0.8.3" → hyperbytedb:v0.8.3). Changing
+                  this field triggers a rolling upgrade.
                 type: string
             type: object
           status:

internal/controller/hyperbytedbbackup_controller.go

@@ -207,10 +207,7 @@ func (r *HyperbytedbBackupReconciler) reconcileCronJob(ctx context.Context, back
 }
 
 func (r *HyperbytedbBackupReconciler) buildBackupJob(backup *hyperbytedbv1alpha1.HyperbytedbBackup, cluster *hyperbytedbv1alpha1.HyperbytedbCluster, name string) *batchv1.Job {
-	image := cluster.Spec.Image
-	if image == "" {
-		image = "hyperbytedb:latest"
-	}
+	image := hyperbytedb.ResolveHyperbytedbImage(cluster)
 
 	s3 := backup.Spec.Destination.S3
 	s3Path := r.buildBackupS3Path(backup)

internal/controller/hyperbytedbcluster_controller.go

@@ -216,9 +216,8 @@ func (r *HyperbytedbClusterReconciler) Reconcile(ctx context.Context, req ctrl.R
 		log.Error(err, "Failed to reconcile HPA")
 	}
 
-	// 11b. Proxy (optional). When disabled (default) we don't reconcile —
-	// any pre-existing proxy resources are left intact so users can clean
-	// them up out-of-band.
+	// 11b. Proxy (optional). Enabled by default; set spec.proxy.enabled=false
+	// to skip reconciliation.
 	if hyperbytedb.ProxyEnabled(cluster) {
 		if err := r.reconcileProxy(ctx, cluster); err != nil {
 			log.Error(err, "Failed to reconcile proxy")
@@ -527,7 +526,10 @@ func (r *HyperbytedbClusterReconciler) reconcilePDB(ctx context.Context, cluster
 
 	pdbName := cluster.Name + "-pdb"
 
-	if replicas < 3 {
+	// Single-node clusters don't need a PDB. For 2+ replicas keep at least
+	// N-1 pods available during voluntary disruption so rolling upgrades and
+	// node drains never take the whole cluster offline at once.
+	if replicas < 2 {
 		existing := &policyv1.PodDisruptionBudget{}
 		err := r.Get(ctx, types.NamespacedName{Name: pdbName, Namespace: cluster.Namespace}, existing)
 		if apierrors.IsNotFound(err) {
@@ -990,6 +992,15 @@ func (r *HyperbytedbClusterReconciler) updateStatus(ctx context.Context, cluster
 	cluster.Status.ReadyReplicas = sts.ReadyReplicas
 	cluster.Status.ConfigHash = configHash
 
+	rollingUpgrade := false
+	curSTS := &appsv1.StatefulSet{}
+	if err := r.Get(ctx, types.NamespacedName{
+		Name: hyperbytedb.StatefulSetName(cluster), Namespace: cluster.Namespace,
+	}, curSTS); err == nil {
+		rollingUpgrade = curSTS.Status.UpdateRevision != "" &&
+			curSTS.Status.UpdateRevision != curSTS.Status.CurrentRevision
+	}
+
 	if sts.ReadyReplicas == sts.SpecReplicas && sts.SpecReplicas > 0 {
 		cluster.Status.Phase = hyperbytedbv1alpha1.ClusterPhaseRunning
 		meta.SetStatusCondition(&cluster.Status.Conditions, metav1.Condition{
@@ -1000,9 +1011,14 @@ func (r *HyperbytedbClusterReconciler) updateStatus(ctx context.Context, cluster
 			LastTransitionTime: metav1.Now(),
 		})
 	} else {
-		if cluster.Status.Phase != hyperbytedbv1alpha1.ClusterPhaseFailed &&
-			cluster.Status.Phase != hyperbytedbv1alpha1.ClusterPhaseScaling &&
-			cluster.Status.Phase != hyperbytedbv1alpha1.ClusterPhaseUpgrading {
+		switch {
+		case cluster.Status.Phase == hyperbytedbv1alpha1.ClusterPhaseFailed:
+			// preserve
+		case cluster.Status.Phase == hyperbytedbv1alpha1.ClusterPhaseScaling:
+			// preserve
+		case rollingUpgrade:
+			cluster.Status.Phase = hyperbytedbv1alpha1.ClusterPhaseUpgrading
+		default:
 			cluster.Status.Phase = hyperbytedbv1alpha1.ClusterPhaseInitializing
 		}
 		meta.SetStatusCondition(&cluster.Status.Conditions, metav1.Condition{

internal/controller/hyperbytedbrestore_controller.go

@@ -246,10 +246,7 @@ func (r *HyperbytedbRestoreReconciler) buildRestoreJob(
 	s3Source *hyperbytedbv1alpha1.S3BackupSpec,
 	ordinal int32,
 ) *batchv1.Job {
-	image := cluster.Spec.Image
-	if image == "" {
-		image = "hyperbytedb:latest"
-	}
+	image := hyperbytedb.ResolveHyperbytedbImage(cluster)
 
 	s3Path := s3Source.Bucket
 	if s3Source.Prefix != "" {

internal/hyperbytedb/client.go

@@ -108,7 +108,9 @@ func (c *Client) DrainNode(ctx context.Context, host string, port int32) error {
 		return err
 	}
 	defer func() { _ = resp.Body.Close() }()
-	if resp.StatusCode != http.StatusOK && resp.StatusCode != http.StatusNoContent {
+	if resp.StatusCode != http.StatusOK &&
+		resp.StatusCode != http.StatusNoContent &&
+		resp.StatusCode != http.StatusAccepted {
 		respBody, _ := io.ReadAll(resp.Body)
 		return fmt.Errorf("drain returned %d: %s", resp.StatusCode, string(respBody))
 	}