Enhance SQL KV store with snowflake ID support and resource version handling

- Updated the parsedKey struct to include ResourceVersionSnowflake for storing the original snowflake ID.
- Implemented snowflakeToMicroseconds function to convert snowflake IDs to microsecond timestamps for database storage.
- Modified parseDataKey function to utilize the new microsecond timestamp for ResourceVersion.
- Enhanced SQL queries in upsertResourceVersion to conditionally update resource versions based on the new timestamp logic.

These changes improve the accuracy and precision of resource versioning in the SQL KV store, enhancing data integrity and management.

packages/grafana-data/src/types/featureToggles.gen.ts

@@ -912,6 +912,10 @@ export interface FeatureToggles {
   */
   unifiedStorageGrpcConnectionPool?: boolean;
   /**
+  * Use KV-backed SQL storage backend instead of direct SQL queries
+  */
+  unifiedStorageKVBackend?: boolean;
+  /**
   * Enables UI functionality to permanently delete alert rules
   * @default true
   */

pkg/services/featuremgmt/registry.go

@@ -1580,6 +1580,14 @@ var (
 			HideFromAdminPage: true,
 			HideFromDocs:      true,
 		},
+		{
+			Name:              "unifiedStorageKVBackend",
+			Description:       "Use KV-backed SQL storage backend instead of direct SQL queries",
+			Stage:             FeatureStageExperimental,
+			Owner:             grafanaSearchAndStorageSquad,
+			HideFromAdminPage: true,
+			HideFromDocs:      true,
+		},
 		{
 			Name:              "alertingRulePermanentlyDelete",
 			Description:       "Enables UI functionality to permanently delete alert rules",

pkg/services/featuremgmt/toggles_gen.csv

@@ -205,6 +205,7 @@ unifiedStorageHistoryPruner,GA,@grafana/search-and-storage,false,false,false
 azureMonitorLogsBuilderEditor,preview,@grafana/partner-datasources,false,false,false
 localeFormatPreference,preview,@grafana/grafana-frontend-platform,false,false,false
 unifiedStorageGrpcConnectionPool,experimental,@grafana/search-and-storage,false,false,false
+unifiedStorageKVBackend,experimental,@grafana/search-and-storage,false,false,false
 alertingRulePermanentlyDelete,GA,@grafana/alerting-squad,false,false,true
 alertingRuleRecoverDeleted,GA,@grafana/alerting-squad,false,false,true
 multiTenantTempCredentials,experimental,@grafana/aws-datasources,false,false,false

pkg/services/featuremgmt/toggles_gen.go

@@ -830,6 +830,10 @@ const (
 	// Enables the unified storage grpc connection pool
 	FlagUnifiedStorageGrpcConnectionPool = "unifiedStorageGrpcConnectionPool"
 
+	// FlagUnifiedStorageKVBackend
+	// Use KV-backed SQL storage backend instead of direct SQL queries
+	FlagUnifiedStorageKVBackend = "unifiedStorageKVBackend"
+
 	// FlagAlertingRulePermanentlyDelete
 	// Enables UI functionality to permanently delete alert rules
 	FlagAlertingRulePermanentlyDelete = "alertingRulePermanentlyDelete"

pkg/services/featuremgmt/toggles_gen.json

@@ -4208,6 +4208,20 @@
         "expression": "true"
       }
     },
+    {
+      "metadata": {
+        "name": "unifiedStorageKVBackend",
+        "resourceVersion": "1763461706359",
+        "creationTimestamp": "2025-11-18T10:28:26Z"
+      },
+      "spec": {
+        "description": "Use KV-backed SQL storage backend instead of direct SQL queries",
+        "stage": "experimental",
+        "codeowner": "@grafana/search-and-storage",
+        "hideFromAdminPage": true,
+        "hideFromDocs": true
+      }
+    },
     {
       "metadata": {
         "name": "unifiedStorageSearch",

pkg/storage/unified/resource/sqlkv.go

@@ -0,0 +1,835 @@
+package resource
+
+import (
+	"bytes"
+	"context"
+	"database/sql"
+	"errors"
+	"fmt"
+	"io"
+	"iter"
+	"strconv"
+	"strings"
+	"time"
+
+	"github.com/bwmarrin/snowflake"
+	"github.com/google/uuid"
+	"github.com/grafana/grafana/pkg/storage/unified/sql/db"
+	"github.com/grafana/grafana/pkg/storage/unified/sql/sqltemplate"
+)
+
+const (
+	sectionData   = "unified/data"
+	sectionEvents = "unified/events"
+)
+
+// sqlKV implements the KV interface using SQL storage
+type sqlKV struct {
+	dbProvider db.DBProvider // Keep reference to prevent GC
+	db         db.DB
+	dialect    sqltemplate.Dialect
+}
+
+// NewSQLKV creates a new SQL-based KV store
+func NewSQLKV(dbProvider db.DBProvider) (KV, error) {
+	if dbProvider == nil {
+		return nil, errors.New("dbProvider is required")
+	}
+
+	// Initialize the database connection
+	ctx := context.Background()
+	dbConn, err := dbProvider.Init(ctx)
+	if err != nil {
+		return nil, fmt.Errorf("initialize DB: %w", err)
+	}
+
+	// Determine the SQL dialect
+	var dialect sqltemplate.Dialect
+	switch dbConn.DriverName() {
+	case "mysql":
+		dialect = sqltemplate.MySQL
+	case "postgres":
+		dialect = sqltemplate.PostgreSQL
+	case "sqlite3", "sqlite":
+		dialect = sqltemplate.SQLite
+	default:
+		return nil, fmt.Errorf("unsupported database driver: %s", dbConn.DriverName())
+	}
+
+	return &sqlKV{
+		dbProvider: dbProvider, // Keep reference to prevent GC from closing the database
+		db:         dbConn,
+		dialect:    dialect,
+	}, nil
+}
+
+// Verify that sqlKV implements KV interface
+var _ KV = &sqlKV{}
+
+// Helper function to build identifiers safely
+func (k *sqlKV) ident(name string) (string, error) {
+	return k.dialect.Ident(name)
+}
+
+// Helper function to get table name for a section
+func (k *sqlKV) getTableName(section string) (string, error) {
+	switch section {
+	case sectionData:
+		return k.ident("resource_history")
+	case sectionEvents:
+		return k.ident("resource_events")
+	default:
+		return "", fmt.Errorf("unsupported section: %s", section)
+	}
+}
+
+// parsedKey represents the components of a key_path for the data section
+// Format: {Group}/{Resource}/{Namespace}/{Name}/{ResourceVersion}~{Action}~{Folder}
+type parsedKey struct {
+	Group                    string
+	Resource                 string
+	Namespace                string
+	Name                     string
+	ResourceVersion          int64 // Microsecond timestamp
+	ResourceVersionSnowflake int64 // Original snowflake ID from key
+	Action                   int   // 1: create, 2: update, 3: delete
+	Folder                   string
+}
+
+// snowflakeToMicroseconds converts a snowflake ID to a unix microsecond timestamp
+// Uses the snowflake library's Time() method which handles the Grafana epoch internally
+func snowflakeToMicroseconds(snowflakeID int64) int64 {
+	// Extract unix milliseconds from snowflake (handles Grafana epoch internally)
+	unixMilliseconds := snowflake.ID(snowflakeID).Time()
+
+	// Extract sequence number (low 12 bits) for sub-millisecond precision
+	sequence := snowflakeID & 0xFFF // 0xFFF = 4095 = 2^12 - 1
+
+	// Convert to unix microseconds: (unix_ms * 1000) + sequence
+	return (unixMilliseconds * 1000) + sequence
+}
+
+// parseDataKey parses a data section key_path
+func parseDataKey(keyPath string) (*parsedKey, error) {
+	// Split by ~ to separate main key from action and folder
+	parts := strings.Split(keyPath, "~")
+	if len(parts) != 3 {
+		return nil, fmt.Errorf("invalid key format: expected 3 parts separated by '~', got %d", len(parts))
+	}
+
+	// Split main key by /
+	mainParts := strings.Split(parts[0], "/")
+	if len(mainParts) != 5 {
+		return nil, fmt.Errorf("invalid key format: expected 5 parts separated by '/', got %d", len(mainParts))
+	}
+
+	// Parse resource version (stored as snowflake ID in key)
+	snowflakeID, err := strconv.ParseInt(mainParts[4], 10, 64)
+	if err != nil {
+		return nil, fmt.Errorf("invalid resource_version: %w", err)
+	}
+
+	// Convert snowflake ID to microsecond timestamp for database storage
+	microseconds := snowflakeToMicroseconds(snowflakeID)
+
+	// Convert action string to int
+	var action int
+	switch parts[1] {
+	case "created":
+		action = 1
+	case "updated":
+		action = 2
+	case "deleted":
+		action = 3
+	default:
+		return nil, fmt.Errorf("invalid action: %s", parts[1])
+	}
+
+	return &parsedKey{
+		Group:                    mainParts[0],
+		Resource:                 mainParts[1],
+		Namespace:                mainParts[2],
+		Name:                     mainParts[3],
+		ResourceVersion:          microseconds, // Microsecond timestamp for DB
+		ResourceVersionSnowflake: snowflakeID,  // Original snowflake ID
+		Action:                   action,
+		Folder:                   parts[2], // May be empty string
+	}, nil
+}
+
+// Get retrieves the value for a key from the store
+func (k *sqlKV) Get(ctx context.Context, section string, key string) (io.ReadCloser, error) {
+	if section == "" {
+		return nil, fmt.Errorf("section is required")
+	}
+	if key == "" {
+		return nil, fmt.Errorf("key is required")
+	}
+
+	tableName, err := k.getTableName(section)
+	if err != nil {
+		return nil, err
+	}
+
+	valueIdent, err := k.ident("value")
+	if err != nil {
+		return nil, fmt.Errorf("invalid column identifier: %w", err)
+	}
+	keyPathIdent, err := k.ident("key_path")
+	if err != nil {
+		return nil, fmt.Errorf("invalid column identifier: %w", err)
+	}
+
+	query := fmt.Sprintf(
+		"SELECT %s FROM %s WHERE %s = %s",
+		valueIdent,
+		tableName,
+		keyPathIdent,
+		k.dialect.ArgPlaceholder(1),
+	)
+
+	// Execute the query
+	var value []byte
+	err = k.db.QueryRowContext(ctx, query, key).Scan(&value)
+	if err != nil {
+		if errors.Is(err, sql.ErrNoRows) {
+			return nil, ErrNotFound
+		}
+		return nil, fmt.Errorf("query failed: %w", err)
+	}
+
+	return io.NopCloser(bytes.NewReader(value)), nil
+}
+
+// BatchGet retrieves multiple values for the given keys from the store
+// Uses a UNION ALL subquery with LEFT JOIN to preserve order and enable streaming
+func (k *sqlKV) BatchGet(ctx context.Context, section string, keys []string) iter.Seq2[KeyValue, error] {
+	if section == "" {
+		return func(yield func(KeyValue, error) bool) {
+			yield(KeyValue{}, fmt.Errorf("section is required"))
+		}
+	}
+
+	if len(keys) == 0 {
+		return func(yield func(KeyValue, error) bool) {
+			// Empty result set - nothing to yield
+		}
+	}
+
+	return func(yield func(KeyValue, error) bool) {
+		tableName, err := k.getTableName(section)
+		if err != nil {
+			yield(KeyValue{}, err)
+			return
+		}
+
+		keyPathIdent, err := k.ident("key_path")
+		if err != nil {
+			yield(KeyValue{}, fmt.Errorf("invalid column identifier: %w", err))
+			return
+		}
+
+		valueIdent, err := k.ident("value")
+		if err != nil {
+			yield(KeyValue{}, fmt.Errorf("invalid column identifier: %w", err))
+			return
+		}
+
+		// Build UNION ALL subquery to preserve key order
+		// SELECT 0 AS idx, ? AS kp UNION ALL SELECT 1, ? UNION ALL ...
+		var unionParts []string
+		var args []interface{}
+		argNum := 1
+
+		for i, key := range keys {
+			if i == 0 {
+				unionParts = append(unionParts, fmt.Sprintf(
+					"SELECT %s AS idx, %s AS kp",
+					k.dialect.ArgPlaceholder(argNum),
+					k.dialect.ArgPlaceholder(argNum+1),
+				))
+			} else {
+				unionParts = append(unionParts, fmt.Sprintf(
+					"UNION ALL SELECT %s, %s",
+					k.dialect.ArgPlaceholder(argNum),
+					k.dialect.ArgPlaceholder(argNum+1),
+				))
+			}
+			args = append(args, i, key)
+			argNum += 2
+		}
+
+		// Build the full query with LEFT JOIN to preserve order
+		// This allows streaming results directly without buffering
+		query := fmt.Sprintf(
+			"SELECT v.idx, t.%s, t.%s FROM (%s) AS v LEFT JOIN %s t ON t.%s = v.kp ORDER BY v.idx",
+			keyPathIdent,
+			valueIdent,
+			strings.Join(unionParts, " "),
+			tableName,
+			keyPathIdent,
+		)
+
+		// Execute the query
+		rows, err := k.db.QueryContext(ctx, query, args...)
+		if err != nil {
+			yield(KeyValue{}, fmt.Errorf("query failed: %w", err))
+			return
+		}
+		defer rows.Close()
+
+		// Stream results directly - no buffering needed!
+		// Results come back in the order specified by idx
+		for rows.Next() {
+			var idx int
+			var keyPath sql.NullString
+			var value []byte
+
+			if err := rows.Scan(&idx, &keyPath, &value); err != nil {
+				yield(KeyValue{}, fmt.Errorf("scan failed: %w", err))
+				return
+			}
+
+			// Skip keys that don't exist (LEFT JOIN returns NULL)
+			if !keyPath.Valid {
+				continue
+			}
+
+			kv := KeyValue{
+				Key:   keyPath.String,
+				Value: io.NopCloser(bytes.NewReader(value)),
+			}
+
+			if !yield(kv, nil) {
+				return
+			}
+		}
+
+		if err := rows.Err(); err != nil {
+			yield(KeyValue{}, fmt.Errorf("rows error: %w", err))
+			return
+		}
+	}
+}
+
+// Keys returns all the keys in the store
+func (k *sqlKV) Keys(ctx context.Context, section string, opt ListOptions) iter.Seq2[string, error] {
+	if section == "" {
+		return func(yield func(string, error) bool) {
+			yield("", fmt.Errorf("section is required"))
+		}
+	}
+
+	return func(yield func(string, error) bool) {
+		tableName, err := k.getTableName(section)
+		if err != nil {
+			yield("", err)
+			return
+		}
+
+		keyPathIdent, err := k.ident("key_path")
+		if err != nil {
+			yield("", fmt.Errorf("invalid column identifier: %w", err))
+			return
+		}
+
+		// Build WHERE clauses
+		var whereClauses []string
+		var args []interface{}
+		argNum := 1
+
+		// Start key (inclusive)
+		if opt.StartKey != "" {
+			whereClauses = append(whereClauses, fmt.Sprintf("%s >= %s", keyPathIdent, k.dialect.ArgPlaceholder(argNum)))
+			args = append(args, opt.StartKey)
+			argNum++
+		}
+
+		// End key (exclusive)
+		if opt.EndKey != "" {
+			whereClauses = append(whereClauses, fmt.Sprintf("%s < %s", keyPathIdent, k.dialect.ArgPlaceholder(argNum)))
+			args = append(args, opt.EndKey)
+			argNum++
+		}
+
+		// Build ORDER BY clause
+		orderBy := "ASC"
+		if opt.Sort == SortOrderDesc {
+			orderBy = "DESC"
+		}
+
+		// Build the query
+		query := fmt.Sprintf(
+			"SELECT %s FROM %s",
+			keyPathIdent,
+			tableName,
+		)
+
+		if len(whereClauses) > 0 {
+			query += " WHERE " + strings.Join(whereClauses, " AND ")
+		}
+
+		query += fmt.Sprintf(" ORDER BY %s %s", keyPathIdent, orderBy)
+
+		if opt.Limit > 0 {
+			query += fmt.Sprintf(" LIMIT %d", opt.Limit)
+		}
+
+		// Execute the query
+		rows, err := k.db.QueryContext(ctx, query, args...)
+		if err != nil {
+			yield("", fmt.Errorf("query failed: %w", err))
+			return
+		}
+		defer rows.Close()
+
+		// Yield each key
+		for rows.Next() {
+			var keyPath string
+			if err := rows.Scan(&keyPath); err != nil {
+				yield("", fmt.Errorf("scan failed: %w", err))
+				return
+			}
+			if !yield(keyPath, nil) {
+				return
+			}
+		}
+
+		if err := rows.Err(); err != nil {
+			yield("", fmt.Errorf("rows error: %w", err))
+			return
+		}
+	}
+}
+
+// Save a new value - returns a WriteCloser to write the value to
+func (k *sqlKV) Save(ctx context.Context, section string, key string) (io.WriteCloser, error) {
+	if section == "" {
+		return nil, fmt.Errorf("section is required")
+	}
+	if key == "" {
+		return nil, fmt.Errorf("key is required")
+	}
+
+	return &sqlWriteCloser{
+		kv:      k,
+		ctx:     ctx,
+		section: section,
+		key:     key,
+		buf:     &bytes.Buffer{},
+		closed:  false,
+	}, nil
+}
+
+// sqlWriteCloser implements io.WriteCloser for SQL KV Save operations
+type sqlWriteCloser struct {
+	kv      *sqlKV
+	ctx     context.Context
+	section string
+	key     string
+	buf     *bytes.Buffer
+	closed  bool
+}
+
+// Write implements io.Writer
+func (w *sqlWriteCloser) Write(p []byte) (int, error) {
+	if w.closed {
+		return 0, fmt.Errorf("write to closed writer")
+	}
+	return w.buf.Write(p)
+}
+
+// Close implements io.Closer - stores the buffered data in SQL
+func (w *sqlWriteCloser) Close() error {
+	if w.closed {
+		return nil
+	}
+	w.closed = true
+
+	value := w.buf.Bytes()
+
+	switch w.section {
+	case sectionEvents:
+		// Simple upsert for events section
+		return w.closeEvents(value)
+	case sectionData:
+		// Complex multi-table transaction for data section
+		return w.closeData(value)
+	default:
+		return fmt.Errorf("unsupported section: %s", w.section)
+	}
+}
+
+// closeEvents handles the simple upsert for the events section
+func (w *sqlWriteCloser) closeEvents(value []byte) error {
+	tableName, err := w.kv.getTableName(w.section)
+	if err != nil {
+		return err
+	}
+
+	keyPathIdent, err := w.kv.ident("key_path")
+	if err != nil {
+		return fmt.Errorf("invalid column identifier: %w", err)
+	}
+
+	valueIdent, err := w.kv.ident("value")
+	if err != nil {
+		return fmt.Errorf("invalid column identifier: %w", err)
+	}
+
+	ph1 := w.kv.dialect.ArgPlaceholder(1)
+	ph2 := w.kv.dialect.ArgPlaceholder(2)
+
+	var query string
+	switch w.kv.dialect.DialectName() {
+	case "postgres":
+		query = fmt.Sprintf(
+			"INSERT INTO %s (%s, %s) VALUES (%s, %s) ON CONFLICT (%s) DO UPDATE SET %s = EXCLUDED.%s",
+			tableName, keyPathIdent, valueIdent, ph1, ph2, keyPathIdent, valueIdent, valueIdent,
+		)
+	case "mysql":
+		query = fmt.Sprintf(
+			"INSERT INTO %s (%s, %s) VALUES (%s, %s) ON DUPLICATE KEY UPDATE %s = VALUES(%s)",
+			tableName, keyPathIdent, valueIdent, ph1, ph2, valueIdent, valueIdent,
+		)
+	case "sqlite":
+		query = fmt.Sprintf(
+			"INSERT INTO %s (%s, %s) VALUES (%s, %s) ON CONFLICT (%s) DO UPDATE SET %s = excluded.%s",
+			tableName, keyPathIdent, valueIdent, ph1, ph2, keyPathIdent, valueIdent, valueIdent,
+		)
+	default:
+		return fmt.Errorf("unsupported dialect: %s", w.kv.dialect.DialectName())
+	}
+
+	_, err = w.kv.db.ExecContext(w.ctx, query, w.key, value)
+	if err != nil {
+		return fmt.Errorf("insert/update failed: %w", err)
+	}
+
+	return nil
+}
+
+// closeData handles the complex multi-table transaction for the data section
+func (w *sqlWriteCloser) closeData(value []byte) error {
+	// Parse the key to extract all fields
+	parsed, err := parseDataKey(w.key)
+	if err != nil {
+		return fmt.Errorf("parse key: %w", err)
+	}
+
+	// Generate a GUID for this write
+	guid := uuid.New().String()
+
+	// Execute all operations in a transaction
+	return w.kv.db.WithTx(w.ctx, nil, func(ctx context.Context, tx db.Tx) error {
+		// 1. Insert/update resource_history
+		if err := w.upsertResourceHistory(ctx, tx, parsed, guid, value); err != nil {
+			return fmt.Errorf("upsert resource_history: %w", err)
+		}
+
+		// 2. Handle resource table based on action
+		if parsed.Action == 3 { // deleted
+			if err := w.deleteResource(ctx, tx, parsed); err != nil {
+				return fmt.Errorf("delete resource: %w", err)
+			}
+		} else { // created or updated
+			if err := w.upsertResource(ctx, tx, parsed, guid, value); err != nil {
+				return fmt.Errorf("upsert resource: %w", err)
+			}
+		}
+
+		// 3. Upsert resource_version table
+		if err := w.upsertResourceVersion(ctx, tx, parsed); err != nil {
+			return fmt.Errorf("upsert resource_version: %w", err)
+		}
+
+		return nil
+	})
+}
+
+// upsertResourceHistory inserts/updates a row in the resource_history table
+func (w *sqlWriteCloser) upsertResourceHistory(ctx context.Context, tx db.Tx, parsed *parsedKey, guid string, value []byte) error {
+	// Build identifiers
+	tableIdent, _ := w.kv.ident("resource_history")
+	guidIdent, _ := w.kv.ident("guid")
+	groupIdent, _ := w.kv.ident("group")
+	resourceIdent, _ := w.kv.ident("resource")
+	namespaceIdent, _ := w.kv.ident("namespace")
+	nameIdent, _ := w.kv.ident("name")
+	rvIdent, _ := w.kv.ident("resource_version")
+	prevRVIdent, _ := w.kv.ident("previous_resource_version")
+	valueIdent, _ := w.kv.ident("value")
+	actionIdent, _ := w.kv.ident("action")
+	folderIdent, _ := w.kv.ident("folder")
+	keyPathIdent, _ := w.kv.ident("key_path")
+
+	// Build placeholders
+	var query string
+	ph := func(n int) string { return w.kv.dialect.ArgPlaceholder(n) }
+
+	switch w.kv.dialect.DialectName() {
+	case "postgres":
+		query = fmt.Sprintf(`
+			INSERT INTO %s (%s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s) 
+			VALUES (%s, %s, %s, %s, %s, %s, %s, 0, %s, %s, %s)
+			ON CONFLICT (%s) DO UPDATE SET %s = EXCLUDED.%s, %s = EXCLUDED.%s`,
+			tableIdent, guidIdent, keyPathIdent, groupIdent, resourceIdent, namespaceIdent, nameIdent,
+			rvIdent, prevRVIdent, valueIdent, actionIdent, folderIdent,
+			ph(1), ph(2), ph(3), ph(4), ph(5), ph(6), ph(7), ph(8), ph(9), ph(10),
+			guidIdent, valueIdent, valueIdent, keyPathIdent, keyPathIdent,
+		)
+	case "mysql", "sqlite":
+		// For MySQL and SQLite, use INSERT OR REPLACE (requires all columns)
+		query = fmt.Sprintf(`
+			REPLACE INTO %s (%s, %s, %s, %s, %s, %s, %s, %s, %s, %s, %s) 
+			VALUES (%s, %s, %s, %s, %s, %s, %s, 0, %s, %s, %s)`,
+			tableIdent, guidIdent, keyPathIdent, groupIdent, resourceIdent, namespaceIdent, nameIdent,
+			rvIdent, prevRVIdent, valueIdent, actionIdent, folderIdent,
+			ph(1), ph(2), ph(3), ph(4), ph(5), ph(6), ph(7), ph(8), ph(9), ph(10),
+		)
+	default:
+		return fmt.Errorf("unsupported dialect: %s", w.kv.dialect.DialectName())
+	}
+
+	_, err := tx.ExecContext(ctx, query,
+		guid, w.key, parsed.Group, parsed.Resource, parsed.Namespace, parsed.Name,
+		parsed.ResourceVersion, value, parsed.Action, parsed.Folder,
+	)
+	return err
+}
+
+// upsertResource inserts/updates a row in the resource table
+func (w *sqlWriteCloser) upsertResource(ctx context.Context, tx db.Tx, parsed *parsedKey, guid string, value []byte) error {
+	// Build identifiers
+	tableIdent, _ := w.kv.ident("resource")
+	guidIdent, _ := w.kv.ident("guid")
+	groupIdent, _ := w.kv.ident("group")
+	resourceIdent, _ := w.kv.ident("resource")
+	namespaceIdent, _ := w.kv.ident("namespace")
+	nameIdent, _ := w.kv.ident("name")
+	rvIdent, _ := w.kv.ident("resource_version")
+	valueIdent, _ := w.kv.ident("value")
+	actionIdent, _ := w.kv.ident("action")
+
+	var query string
+	ph := func(n int) string { return w.kv.dialect.ArgPlaceholder(n) }
+
+	switch w.kv.dialect.DialectName() {
+	case "postgres":
+		query = fmt.Sprintf(`
+			INSERT INTO %s (%s, %s, %s, %s, %s, %s, %s, %s)
+			VALUES (%s, %s, %s, %s, %s, %s, %s, %s)
+			ON CONFLICT (%s, %s, %s, %s) DO UPDATE SET 
+				%s = EXCLUDED.%s, %s = EXCLUDED.%s, %s = EXCLUDED.%s, %s = EXCLUDED.%s`,
+			tableIdent, guidIdent, groupIdent, resourceIdent, namespaceIdent, nameIdent, rvIdent, valueIdent, actionIdent,
+			ph(1), ph(2), ph(3), ph(4), ph(5), ph(6), ph(7), ph(8),
+			namespaceIdent, groupIdent, resourceIdent, nameIdent,
+			guidIdent, guidIdent, rvIdent, rvIdent, valueIdent, valueIdent, actionIdent, actionIdent,
+		)
+	case "mysql", "sqlite":
+		query = fmt.Sprintf(`
+			REPLACE INTO %s (%s, %s, %s, %s, %s, %s, %s, %s)
+			VALUES (%s, %s, %s, %s, %s, %s, %s, %s)`,
+			tableIdent, guidIdent, groupIdent, resourceIdent, namespaceIdent, nameIdent, rvIdent, valueIdent, actionIdent,
+			ph(1), ph(2), ph(3), ph(4), ph(5), ph(6), ph(7), ph(8),
+		)
+	default:
+		return fmt.Errorf("unsupported dialect: %s", w.kv.dialect.DialectName())
+	}
+
+	_, err := tx.ExecContext(ctx, query,
+		guid, parsed.Group, parsed.Resource, parsed.Namespace, parsed.Name,
+		parsed.ResourceVersion, value, parsed.Action,
+	)
+	return err
+}
+
+// deleteResource deletes a row from the resource table
+func (w *sqlWriteCloser) deleteResource(ctx context.Context, tx db.Tx, parsed *parsedKey) error {
+	tableIdent, _ := w.kv.ident("resource")
+	groupIdent, _ := w.kv.ident("group")
+	resourceIdent, _ := w.kv.ident("resource")
+	namespaceIdent, _ := w.kv.ident("namespace")
+	nameIdent, _ := w.kv.ident("name")
+
+	ph := func(n int) string { return w.kv.dialect.ArgPlaceholder(n) }
+
+	query := fmt.Sprintf(`
+		DELETE FROM %s WHERE %s = %s AND %s = %s AND %s = %s AND %s = %s`,
+		tableIdent, groupIdent, ph(1), resourceIdent, ph(2), namespaceIdent, ph(3), nameIdent, ph(4),
+	)
+
+	_, err := tx.ExecContext(ctx, query, parsed.Group, parsed.Resource, parsed.Namespace, parsed.Name)
+	return err
+}
+
+// upsertResourceVersion inserts/updates the resource_version table
+func (w *sqlWriteCloser) upsertResourceVersion(ctx context.Context, tx db.Tx, parsed *parsedKey) error {
+	tableIdent, _ := w.kv.ident("resource_version")
+	groupIdent, _ := w.kv.ident("group")
+	resourceIdent, _ := w.kv.ident("resource")
+	rvIdent, _ := w.kv.ident("resource_version")
+
+	ph := func(n int) string { return w.kv.dialect.ArgPlaceholder(n) }
+
+	var query string
+	switch w.kv.dialect.DialectName() {
+	case "postgres":
+		// Only update if new resource_version is greater than existing
+		query = fmt.Sprintf(`
+			INSERT INTO %s (%s, %s, %s) VALUES (%s, %s, %s)
+			ON CONFLICT (%s, %s) DO UPDATE SET %s = EXCLUDED.%s
+			WHERE EXCLUDED.%s > %s.%s`,
+			tableIdent, groupIdent, resourceIdent, rvIdent, ph(1), ph(2), ph(3),
+			groupIdent, resourceIdent, rvIdent, rvIdent,
+			rvIdent, tableIdent, rvIdent,
+		)
+	case "sqlite":
+		// SQLite supports WHERE clause in ON CONFLICT DO UPDATE
+		query = fmt.Sprintf(`
+			INSERT INTO %s (%s, %s, %s) VALUES (%s, %s, %s)
+			ON CONFLICT (%s, %s) DO UPDATE SET %s = EXCLUDED.%s
+			WHERE EXCLUDED.%s > %s.%s`,
+			tableIdent, groupIdent, resourceIdent, rvIdent, ph(1), ph(2), ph(3),
+			groupIdent, resourceIdent, rvIdent, rvIdent,
+			rvIdent, tableIdent, rvIdent,
+		)
+	case "mysql":
+		// MySQL uses ON DUPLICATE KEY UPDATE with conditional
+		query = fmt.Sprintf(`
+			INSERT INTO %s (%s, %s, %s) VALUES (%s, %s, %s)
+			ON DUPLICATE KEY UPDATE %s = IF(VALUES(%s) > %s, VALUES(%s), %s)`,
+			tableIdent, groupIdent, resourceIdent, rvIdent, ph(1), ph(2), ph(3),
+			rvIdent, rvIdent, rvIdent, rvIdent, rvIdent,
+		)
+	default:
+		return fmt.Errorf("unsupported dialect: %s", w.kv.dialect.DialectName())
+	}
+
+	_, err := tx.ExecContext(ctx, query, parsed.Group, parsed.Resource, parsed.ResourceVersion)
+	return err
+}
+
+// Delete a value
+func (k *sqlKV) Delete(ctx context.Context, section string, key string) error {
+	if section == "" {
+		return fmt.Errorf("section is required")
+	}
+	if key == "" {
+		return fmt.Errorf("key is required")
+	}
+
+	tableName, err := k.getTableName(section)
+	if err != nil {
+		return err
+	}
+
+	keyPathIdent, err := k.ident("key_path")
+	if err != nil {
+		return fmt.Errorf("invalid column identifier: %w", err)
+	}
+
+	// First check if key exists (to return ErrNotFound if missing)
+	checkQuery := fmt.Sprintf(
+		"SELECT 1 FROM %s WHERE %s = %s",
+		tableName,
+		keyPathIdent,
+		k.dialect.ArgPlaceholder(1),
+	)
+
+	var exists int
+	err = k.db.QueryRowContext(ctx, checkQuery, key).Scan(&exists)
+	if err != nil {
+		if errors.Is(err, sql.ErrNoRows) {
+			return ErrNotFound
+		}
+		return fmt.Errorf("check existence failed: %w", err)
+	}
+
+	// Delete the key
+	deleteQuery := fmt.Sprintf(
+		"DELETE FROM %s WHERE %s = %s",
+		tableName,
+		keyPathIdent,
+		k.dialect.ArgPlaceholder(1),
+	)
+
+	_, err = k.db.ExecContext(ctx, deleteQuery, key)
+	if err != nil {
+		return fmt.Errorf("delete failed: %w", err)
+	}
+
+	return nil
+}
+
+// BatchDelete removes multiple keys from the store
+func (k *sqlKV) BatchDelete(ctx context.Context, section string, keys []string) error {
+	if section == "" {
+		return fmt.Errorf("section is required")
+	}
+
+	if len(keys) == 0 {
+		return nil // Nothing to delete
+	}
+
+	tableName, err := k.getTableName(section)
+	if err != nil {
+		return err
+	}
+
+	keyPathIdent, err := k.ident("key_path")
+	if err != nil {
+		return fmt.Errorf("invalid column identifier: %w", err)
+	}
+
+	// Build IN clause placeholders
+	placeholders := make([]string, len(keys))
+	args := make([]interface{}, len(keys))
+	for i, key := range keys {
+		placeholders[i] = k.dialect.ArgPlaceholder(i + 1)
+		args[i] = key
+	}
+
+	// Build the query
+	query := fmt.Sprintf(
+		"DELETE FROM %s WHERE %s IN (%s)",
+		tableName,
+		keyPathIdent,
+		strings.Join(placeholders, ", "),
+	)
+
+	// Execute the query (idempotent - non-existent keys are silently ignored)
+	_, err = k.db.ExecContext(ctx, query, args...)
+	if err != nil {
+		return fmt.Errorf("batch delete failed: %w", err)
+	}
+
+	return nil
+}
+
+// UnixTimestamp returns the current time in seconds since Epoch
+func (k *sqlKV) UnixTimestamp(ctx context.Context) (int64, error) {
+	return time.Now().Unix(), nil
+}
+
+// Ping checks if the database connection is alive
+func (k *sqlKV) Ping(ctx context.Context) error {
+	if k.db == nil {
+		return fmt.Errorf("database connection is nil")
+	}
+	return k.db.PingContext(ctx)
+}
+
+// checkDB verifies the database connection is still valid before operations
+func (k *sqlKV) checkDB() error {
+	if k.db == nil {
+		return fmt.Errorf("database connection is nil")
+	}
+	// Quick ping to verify connection is alive
+	ctx, cancel := context.WithTimeout(context.Background(), time.Second)
+	defer cancel()
+	if err := k.db.PingContext(ctx); err != nil {
+		return fmt.Errorf("database connection is not healthy: %w", err)
+	}
+	return nil
+}

pkg/storage/unified/resource/storage_backend.go

@@ -66,9 +66,15 @@ type kvStorageBackend struct {
 	withExperimentalClusterScope bool
 	//tracer        trace.Tracer
 	//reg           prometheus.Registerer
+
+	// lifecycle management
+	ctx    context.Context
+	cancel context.CancelFunc
 }
 
 var _ StorageBackend = &kvStorageBackend{}
+var _ LifecycleHooks = &kvStorageBackend{}
+var _ resourcepb.DiagnosticsServer = &kvStorageBackend{}
 
 type KVBackendOptions struct {
 	KvStore                      KV
@@ -81,7 +87,6 @@ type KVBackendOptions struct {
 }
 
 func NewKVStorageBackend(opts KVBackendOptions) (StorageBackend, error) {
-	ctx := context.Background()
 	kv := opts.KvStore
 
 	s, err := snowflake.NewNode(rand.Int64N(1024))
@@ -100,6 +105,9 @@ func NewKVStorageBackend(opts KVBackendOptions) (StorageBackend, error) {
 		eventPruningInterval = defaultEventPruningInterval
 	}
 
+	// Create a cancellable context for lifecycle management
+	ctx, cancel := context.WithCancel(context.Background())
+
 	backend := &kvStorageBackend{
 		kv:                           kv,
 		dataStore:                    newDataStore(kv),
@@ -111,18 +119,58 @@ func NewKVStorageBackend(opts KVBackendOptions) (StorageBackend, error) {
 		eventRetentionPeriod:         eventRetentionPeriod,
 		eventPruningInterval:         eventPruningInterval,
 		withExperimentalClusterScope: opts.WithExperimentalClusterScope,
+		ctx:                          ctx,
+		cancel:                       cancel,
 	}
-	err = backend.initPruner(ctx)
-	if err != nil {
-		return nil, fmt.Errorf("failed to initialize pruner: %w", err)
-	}
-
-	// Start the event cleanup background job
-	go backend.runCleanupOldEvents(ctx)
 
 	return backend, nil
 }
 
+// Init implements LifecycleHooks
+func (k *kvStorageBackend) Init(ctx context.Context) error {
+	// Initialize the pruner
+	if err := k.initPruner(ctx); err != nil {
+		return fmt.Errorf("failed to initialize pruner: %w", err)
+	}
+
+	// Start the event cleanup background job using the backend's lifecycle context
+	go k.runCleanupOldEvents(k.ctx)
+
+	// Start the pruner if it was configured
+	if k.historyPruner != nil {
+		k.historyPruner.Start(k.ctx)
+	}
+
+	return nil
+}
+
+// Stop implements LifecycleHooks
+func (k *kvStorageBackend) Stop(ctx context.Context) error {
+	// Cancel the context to stop background goroutines
+	// This will stop both runCleanupOldEvents and the pruner (via debouncer)
+	k.cancel()
+	return nil
+}
+
+// IsHealthy implements DiagnosticsServer
+func (k *kvStorageBackend) IsHealthy(ctx context.Context, _ *resourcepb.HealthCheckRequest) (*resourcepb.HealthCheckResponse, error) {
+	// Check if the underlying KV store supports Ping (e.g., sqlKV)
+	type pinger interface {
+		Ping(context.Context) error
+	}
+	if p, ok := k.kv.(pinger); ok {
+		if err := p.Ping(ctx); err != nil {
+			return nil, fmt.Errorf("KV store health check failed: %w", err)
+		}
+	}
+	return &resourcepb.HealthCheckResponse{Status: resourcepb.HealthCheckResponse_SERVING}, nil
+}
+
+// Read implements DiagnosticsServer
+func (k *kvStorageBackend) Read(ctx context.Context, req *resourcepb.ReadRequest) (*resourcepb.ReadResponse, error) {
+	return nil, ErrNotImplementedYet
+}
+
 // runCleanupOldEvents starts a background goroutine that periodically cleans up old events
 func (k *kvStorageBackend) runCleanupOldEvents(ctx context.Context) {
 	// Run cleanup every hour
@@ -217,7 +265,6 @@ func (k *kvStorageBackend) initPruner(ctx context.Context) error {
 	}
 
 	k.historyPruner = pruner
-	k.historyPruner.Start(ctx)
 	return nil
 }
 

pkg/storage/unified/sql/data/resource_history_update_rv.sql

@@ -1,11 +1,45 @@
 UPDATE {{ .Ident "resource_history" }}
-SET {{ .Ident "resource_version" }} = (
+SET 
+    {{ .Ident "resource_version" }} = (
     CASE
     {{ range $guid, $rv := .GUIDToRV }}
     WHEN {{ $.Ident "guid" }} = {{ $.Arg $guid }} THEN CAST({{ $.Arg $rv }} AS {{ if eq $.DialectName "postgres" }}BIGINT{{ else }}SIGNED{{ end }})
     {{ end }}
     END
+    ),
+    {{ .Ident "key_path" }} = {{ if eq .DialectName "sqlite" -}}
+        {{ .Ident "group" }} || CHAR(47) || {{ .Ident "resource" }} || CHAR(47) || {{ .Ident "namespace" }} || CHAR(47) || {{ .Ident "name" }} || CHAR(47) ||
+            CAST((CASE
+                {{- range $guid, $rv := .GUIDToRV }}
+                WHEN {{ $.Ident "guid" }} = {{ $.Arg $guid }} THEN ((({{ $.Arg $rv }} / 1000) - 1288834974657) * 4194304) + ({{ $.Arg $rv }} % 1000)
+                {{- end }}
+            END) AS TEXT) || CHAR(126) ||
+            CASE {{ .Ident "action" }}
+                WHEN 1 THEN 'created'
+                WHEN 2 THEN 'updated'
+                WHEN 3 THEN 'deleted'
+                ELSE 'unknown'
+            END || CHAR(126) || COALESCE({{ .Ident "folder" }}, '')
+        {{- else -}}
+        CONCAT(
+            {{ .Ident "group" }}, CHAR(47),
+            {{ .Ident "resource" }}, CHAR(47),
+            {{ .Ident "namespace" }}, CHAR(47),
+            {{ .Ident "name" }}, CHAR(47),
+            CAST((CASE
+                {{- range $guid, $rv := .GUIDToRV }}
+                WHEN {{ $.Ident "guid" }} = {{ $.Arg $guid }} THEN ((({{ $.Arg $rv }} DIV 1000) - 1288834974657) * 4194304) + ({{ $.Arg $rv }} MOD 1000)
+                {{- end }}
+            END) AS {{ if eq .DialectName "postgres" }}TEXT{{ else }}CHAR{{ end }}), CHAR(126),
+            CASE {{ .Ident "action" }}
+                WHEN 1 THEN 'created'
+                WHEN 2 THEN 'updated'
+                WHEN 3 THEN 'deleted'
+                ELSE 'unknown'
+            END, CHAR(126),
+            COALESCE({{ .Ident "folder" }}, '')
         )
+        {{- end }}
 WHERE {{ .Ident "guid" }} IN (
     {{$first := true}}
     {{ range $guid, $rv := .GUIDToRV }}{{if $first}}{{$first = false}}{{else}}, {{end}}{{ $.Arg $guid }}{{ end }}

pkg/storage/unified/sql/db/migrations/resource_mig.go

@@ -3,7 +3,9 @@ package migrations
 import (
 	"fmt"
 
+	"github.com/grafana/grafana/pkg/infra/log"
 	"github.com/grafana/grafana/pkg/services/sqlstore/migrator"
+	"github.com/grafana/grafana/pkg/util/xorm"
 )
 
 func initResourceTables(mg *migrator.Migrator) string {
@@ -185,5 +187,170 @@ func initResourceTables(mg *migrator.Migrator) string {
 		Name: "UQE_resource_last_import_time_last_import_time",
 	}))
 
+	// TODO: Do we want the value to be MEDIUMTEXT ?
+	// TODO: What's the best name for the key_path column?
+
+	// Add key_path column to resource_history for KV interface
+	mg.AddMigration("Add key_path column to resource_history", migrator.NewAddColumnMigration(resource_history_table, &migrator.Column{
+		Name: "key_path", Type: migrator.DB_NVarchar, Length: 2048, Nullable: true,
+	}))
+
+	// Backfill key_path column in resource_history
+	mg.AddMigration("Backfill key_path column in resource_history", &resourceHistoryKeyBackfillMigrator{})
+
+	// Note: key_path remains nullable because the write pattern is:
+	// 1. INSERT (key_path = NULL)
+	// 2. UPDATE (key_path = actual value after RV allocation)
+
+	// Add index on key_path column
+	mg.AddMigration("Add index on key_path column in resource_history", migrator.NewAddIndexMigration(resource_history_table, &migrator.Index{
+		Name: "IDX_resource_history_key_path",
+		Cols: []string{"key_path"},
+		Type: migrator.IndexType,
+	}))
+
+	// Create resource_events table for KV interface
+	resource_events_table := migrator.Table{
+		Name: "resource_events",
+		Columns: []*migrator.Column{
+			{Name: "key_path", Type: migrator.DB_NVarchar, Length: 2048, Nullable: false, IsPrimaryKey: true},
+			{Name: "value", Type: migrator.DB_MediumText, Nullable: false},
+		},
+	}
+	mg.AddMigration("create table "+resource_events_table.Name, migrator.NewAddTableMigration(resource_events_table))
+
 	return marker
 }
+
+// resourceHistoryKeyBackfillMigrator backfills the key_path column in resource_history table
+// It processes rows in batches to reduce lock duration and avoid timeouts on large tables
+type resourceHistoryKeyBackfillMigrator struct {
+	migrator.MigrationBase
+}
+
+func (m *resourceHistoryKeyBackfillMigrator) SQL(dialect migrator.Dialect) string {
+	return "Backfill key_path column in resource_history using pattern: {Group}/{Resource}/{Namespace}/{Name}/{ResourceVersion}~{Action}~{Folder}"
+}
+
+func (m *resourceHistoryKeyBackfillMigrator) Exec(sess *xorm.Session, mg *migrator.Migrator) error {
+	dialect := mg.Dialect.DriverName()
+	logger := log.New("resource-history-key-backfill")
+
+	// TODO: Verify the RV to Snowflake ID conversion is correct.
+
+	// Snowflake ID epoch in milliseconds (2010-11-04T01:42:54.657Z)
+	const epochMs = 1288834974657
+	const batchSize = 1000 // Process 1000 rows at a time
+
+	// Count total rows to backfill
+	totalCount, err := sess.Table("resource_history").Where("key_path IS NULL").Count()
+	if err != nil {
+		return fmt.Errorf("failed to count rows: %w", err)
+	}
+
+	if totalCount == 0 {
+		logger.Info("No rows to backfill")
+		return nil
+	}
+
+	logger.Info("Starting key_path backfill", "total_rows", totalCount)
+
+	// Build the SQL query based on the database dialect
+	var updateSQL string
+
+	switch dialect {
+	case "mysql":
+		updateSQL = `
+			UPDATE resource_history 
+			SET key_path = CONCAT(
+				` + "`group`" + `, '/', 
+				` + "`resource`" + `, '/', 
+				` + "`namespace`" + `, '/', 
+				` + "`name`" + `, '/', 
+				CAST((((` + "`resource_version`" + ` DIV 1000) - ?) * 4194304) + (` + "`resource_version`" + ` MOD 1000) AS CHAR), '~', 
+				CASE ` + "`action`" + `
+					WHEN 1 THEN 'created' 
+					WHEN 2 THEN 'updated' 
+					WHEN 3 THEN 'deleted' 
+					ELSE 'unknown'
+				END, '~', 
+				COALESCE(` + "`folder`" + `, '')
+			)
+			WHERE key_path IS NULL
+			LIMIT ?
+		`
+	case "postgres":
+		updateSQL = `
+			UPDATE resource_history 
+			SET key_path = CONCAT(
+				"group", '/', 
+				"resource", '/', 
+				"namespace", '/', 
+				"name", '/', 
+				CAST((((resource_version / 1000) - $1) * 4194304) + (resource_version % 1000) AS BIGINT), '~', 
+				CASE "action" 
+					WHEN 1 THEN 'created' 
+					WHEN 2 THEN 'updated' 
+					WHEN 3 THEN 'deleted' 
+					ELSE 'unknown'
+				END, '~', 
+				COALESCE("folder", '')
+			)
+			WHERE guid IN (
+				SELECT guid FROM resource_history 
+				WHERE key_path IS NULL 
+				LIMIT $2
+			)
+		`
+	case "sqlite3":
+		updateSQL = `
+			UPDATE resource_history 
+			SET key_path = 
+				"group" || '/' || 
+				resource || '/' || 
+				namespace || '/' || 
+				name || '/' || 
+				CAST((((resource_version / 1000) - ?) * 4194304) + (resource_version % 1000) AS TEXT) || '~' || 
+				CASE action 
+					WHEN 1 THEN 'created' 
+					WHEN 2 THEN 'updated' 
+					WHEN 3 THEN 'deleted' 
+					ELSE 'unknown'
+				END || '~' || 
+				COALESCE(folder, '')
+			WHERE guid IN (
+				SELECT guid FROM resource_history 
+				WHERE key_path IS NULL 
+				LIMIT ?
+			)
+		`
+	default:
+		return fmt.Errorf("unsupported database dialect: %s", dialect)
+	}
+
+	// Process in batches
+	processed := int64(0)
+	for {
+		result, err := sess.Exec(updateSQL, epochMs, batchSize)
+		if err != nil {
+			return fmt.Errorf("failed to update batch: %w", err)
+		}
+
+		rowsAffected, err := result.RowsAffected()
+		if err != nil {
+			return fmt.Errorf("failed to get rows affected: %w", err)
+		}
+
+		processed += rowsAffected
+		logger.Info("Backfill progress", "processed", processed, "total", totalCount,
+			"percent", fmt.Sprintf("%.1f%%", float64(processed)/float64(totalCount)*100))
+
+		// If we updated fewer rows than batch size, we're done
+		if rowsAffected < int64(batchSize) {
+			break
+		}
+	}
+
+	logger.Info("Backfill completed", "total_processed", processed)
+	return nil
+}

pkg/storage/unified/sql/server.go

@@ -19,6 +19,7 @@ import (
 	"github.com/grafana/grafana/pkg/services/sqlstore/migrator"
 	"github.com/grafana/grafana/pkg/setting"
 	"github.com/grafana/grafana/pkg/storage/unified/resource"
+	"github.com/grafana/grafana/pkg/storage/unified/resourcepb"
 	"github.com/grafana/grafana/pkg/storage/unified/sql/db/dbimpl"
 )
 
@@ -101,6 +102,30 @@ func NewResourceServer(opts ServerOptions) (resource.ResourceServer, error) {
 		//nolint:staticcheck // not yet migrated to OpenFeature
 		withPruner := opts.Features.IsEnabledGlobally(featuremgmt.FlagUnifiedStorageHistoryPruner)
 
+		// Check if KV backend is enabled via feature flag
+		//nolint:staticcheck // not yet migrated to OpenFeature
+		if opts.Features.IsEnabledGlobally(featuremgmt.FlagUnifiedStorageKVBackend) {
+			// Create SQL KV instance
+			sqlKV, err := resource.NewSQLKV(eDB)
+			if err != nil {
+				return nil, fmt.Errorf("create SQL KV: %w", err)
+			}
+
+			// Use existing KV storage backend (already implements StorageBackend interface)
+			kvBackend, err := resource.NewKVStorageBackend(resource.KVBackendOptions{
+				KvStore:    sqlKV,
+				WithPruner: withPruner,
+				Tracer:     opts.Tracer,
+				Reg:        opts.Reg,
+			})
+			if err != nil {
+				return nil, fmt.Errorf("create KV backend: %w", err)
+			}
+			serverOptions.Backend = kvBackend
+			serverOptions.Lifecycle = kvBackend.(resource.LifecycleHooks)
+			serverOptions.Diagnostics = kvBackend.(resourcepb.DiagnosticsServer)
+		} else {
+			// Use existing SQL backend
 			backend, err := NewBackend(BackendOptions{
 				DBProvider:           eDB,
 				Tracer:               opts.Tracer,
@@ -117,6 +142,7 @@ func NewResourceServer(opts ServerOptions) (resource.ResourceServer, error) {
 			serverOptions.Diagnostics = backend
 			serverOptions.Lifecycle = backend
 		}
+	}
 
 	serverOptions.Search = opts.SearchOptions
 	serverOptions.IndexMetrics = opts.IndexMetrics