Merge remote-tracking branch 'origin/main' into refactor/permissions-manager

management/server/account_test.go

@@ -3517,7 +3517,13 @@ func buildTestManager(t testing.TB, store store.Store, nmdataStore *networkmapdb
 
 	eventStore := &activity.InMemoryEventStore{}
 
-	metrics, err := telemetry.NewDefaultAppMetrics(context.Background())
+	// Everything built here watches this context; cancelling it on cleanup stops
+	// the metrics flushers, caches and controllers instead of leaking them for
+	// the rest of the package run.
+	ctx, cancel := context.WithCancel(context.Background())
+	t.Cleanup(cancel)
+
+	metrics, err := telemetry.NewDefaultAppMetrics(ctx)
 	if err != nil {
 		return nil, nil, err
 	}
@@ -3536,8 +3542,6 @@ func buildTestManager(t testing.TB, store store.Store, nmdataStore *networkmapdb
 		Return(nil).
 		AnyTimes()
 
-	ctx := context.Background()
-
 	cacheStore, err := cache.NewStore(ctx, 100*time.Millisecond, 300*time.Millisecond, 100)
 	if err != nil {
 		return nil, nil, err

management/server/agentnetwork_budgetrule_realstack_test.go

@@ -89,6 +89,7 @@ func TestAgentNetwork_UpdateSettings_PreservesImmutableAndTogglesCollection(t *t
 
 	// Bootstrap is an explicit settings create; providers have no settings
 	// side effects anymore.
+	seedPrivateProxyCluster(t, am.Store, clusterAddr)
 	before, err := mgr.CreateSettings(ctx, adminUserID, agenttypes.DefaultSettings(accountID), clusterAddr, "")
 	require.NoError(t, err, "CreateSettings must bootstrap the row")
 	require.Equal(t, clusterAddr, before.ProxyAddress, "proxy address pinned at bootstrap")

management/server/agentnetwork_realstack_test.go

@@ -14,6 +14,7 @@ import (
 	"github.com/netbirdio/netbird/management/internals/modules/agentnetwork"
 	agenttypes "github.com/netbirdio/netbird/management/internals/modules/agentnetwork/types"
 	"github.com/netbirdio/netbird/management/internals/modules/permissions"
+	rpproxy "github.com/netbirdio/netbird/management/internals/modules/reverseproxy/proxy"
 	nbpeer "github.com/netbirdio/netbird/management/server/peer"
 	"github.com/netbirdio/netbird/management/server/store"
 	"github.com/netbirdio/netbird/management/server/types"
@@ -92,6 +93,7 @@ func TestAgentNetwork_ProviderCRUD_FansOutToProxyAndClientPeers(t *testing.T) {
 	// UpdateAccountPeers, which is the path under test.
 	agentMgr := agentnetwork.NewManager(am.Store, permissions.NewManager(am.Store), am, nil)
 
+	seedPrivateProxyCluster(t, am.Store, clusterAddr)
 	_, err = agentMgr.CreateSettings(ctx, adminUserID, agenttypes.DefaultSettings(accountID), clusterAddr, "")
 	require.NoError(t, err, "CreateSettings must bootstrap the endpoint")
 	// The bootstrap itself reconciles and queues updates on both channels;
@@ -222,3 +224,22 @@ func synthZoneRData(sync *nbproto.SyncResponse, clusterAddr, fqdn string) string
 	}
 	return ""
 }
+
+// seedPrivateProxyCluster registers a connected proxy with private capabilities in a
+// netbird client for clusterAddr, matching what a real deployment looks like
+// when the account bootstraps: the agent-network gateway service is always
+// private, so its cluster has to be one that can serve private services.
+func seedPrivateProxyCluster(t *testing.T, st store.Store, clusterAddr string) {
+	t.Helper()
+	private := true
+	now := time.Now().UTC()
+	require.NoError(t, st.SaveProxy(context.Background(), &rpproxy.Proxy{
+		ID:             "agent-net-proxy-" + clusterAddr,
+		SessionID:      "agent-net-session",
+		ClusterAddress: clusterAddr,
+		LastSeen:       now,
+		ConnectedAt:    &now,
+		Status:         rpproxy.StatusConnected,
+		Capabilities:   rpproxy.Capabilities{Private: &private},
+	}), "seeding the proxy cluster must succeed")
+}

management/server/store/sql_store.go

@@ -3154,7 +3154,12 @@ func NewMysqlStore(ctx context.Context, dsn string, metrics telemetry.AppMetrics
 		return nil, err
 	}
 
-	return NewSqlStore(ctx, db, types.MysqlStoreEngine, metrics, skipMigration)
+	store, err := NewSqlStore(ctx, db, types.MysqlStoreEngine, metrics, skipMigration)
+	if err != nil {
+		closeGormDB(db)
+		return nil, err
+	}
+	return store, nil
 }
 
 func getGormConfig() *gorm.Config {
@@ -3213,23 +3218,20 @@ func NewSqliteStoreFromFileStore(ctx context.Context, fileStore *FileStore, data
 
 // NewPostgresqlStoreFromSqlStore restores a store from SqlStore and stores Postgres DB.
 func NewPostgresqlStoreFromSqlStore(ctx context.Context, sqliteStore *SqlStore, dsn string, metrics telemetry.AppMetrics) (*SqlStore, error) {
-	store, err := NewPostgresqlStoreForTests(ctx, dsn, metrics, false)
+	return newPostgresqlStoreFromSqlStore(ctx, sqliteStore, dsn, metrics, false)
+}
+
+func newPostgresqlStoreFromSqlStore(ctx context.Context, sqliteStore *SqlStore, dsn string, metrics telemetry.AppMetrics, skipMigration bool) (*SqlStore, error) {
+	store, err := NewPostgresqlStoreForTests(ctx, dsn, metrics, skipMigration)
 	if err != nil {
 		return nil, err
 	}
 
-	err = store.SaveInstallationID(ctx, sqliteStore.GetInstallationID())
-	if err != nil {
+	if err := seedFromSqliteStore(ctx, store, sqliteStore); err != nil {
+		closeStore(ctx, store)
 		return nil, err
 	}
 
-	for _, account := range sqliteStore.GetAllAccounts(ctx) {
-		err := store.SaveAccount(ctx, account)
-		if err != nil {
-			return nil, err
-		}
-	}
-
 	if err = copyZonesAndRecords(sqliteStore, store); err != nil {
 		return nil, err
 	}
@@ -3245,11 +3247,14 @@ func NewPostgresqlStoreForTests(ctx context.Context, dsn string, metrics telemet
 	}
 	pool, err := connectToPgDbForTests(context.Background(), dsn)
 	if err != nil {
+		closeGormDB(db)
 		return nil, err
 	}
 	store, err := NewSqlStore(ctx, db, types.PostgresStoreEngine, metrics, skipMigration)
 	if err != nil {
+		// Release the sessions, or the caller cannot drop the database.
 		pool.Close()
+		closeGormDB(db)
 		return nil, err
 	}
 	store.pool = pool
@@ -3283,22 +3288,42 @@ func connectToPgDbForTests(ctx context.Context, dsn string) (*pgxpool.Pool, erro
 
 // NewMysqlStoreFromSqlStore restores a store from SqlStore and stores MySQL DB.
 func NewMysqlStoreFromSqlStore(ctx context.Context, sqliteStore *SqlStore, dsn string, metrics telemetry.AppMetrics) (*SqlStore, error) {
-	store, err := NewMysqlStore(ctx, dsn, metrics, false)
-	if err != nil {
-		return nil, err
-	}
+	return newMysqlStoreFromSqlStore(ctx, sqliteStore, dsn, metrics, false)
+}
 
-	err = store.SaveInstallationID(ctx, sqliteStore.GetInstallationID())
-	if err != nil {
-		return nil, err
+// seedFromSqliteStore copies the installation ID and the accounts of the
+// sqlite seed store into a freshly created engine store.
+func seedFromSqliteStore(ctx context.Context, store, sqliteStore *SqlStore) error {
+	if err := store.SaveInstallationID(ctx, sqliteStore.GetInstallationID()); err != nil {
+		return err
 	}
-
 	for _, account := range sqliteStore.GetAllAccounts(ctx) {
-		err := store.SaveAccount(ctx, account)
-		if err != nil {
-			return nil, err
+		if err := store.SaveAccount(ctx, account); err != nil {
+			return err
 		}
 	}
+	return nil
+}
+
+// closeStore releases a store that is not handed to the caller, so a failed
+// seed does not leak its connection and pool.
+func closeStore(ctx context.Context, store *SqlStore) {
+	store.Close(ctx)
+	if store.pool != nil {
+		store.pool.Close()
+	}
+}
+
+func newMysqlStoreFromSqlStore(ctx context.Context, sqliteStore *SqlStore, dsn string, metrics telemetry.AppMetrics, skipMigration bool) (*SqlStore, error) {
+	store, err := NewMysqlStore(ctx, dsn, metrics, skipMigration)
+	if err != nil {
+		return nil, err
+	}
+
+	if err := seedFromSqliteStore(ctx, store, sqliteStore); err != nil {
+		closeStore(ctx, store)
+		return nil, err
+	}
 
 	if err = copyZonesAndRecords(sqliteStore, store); err != nil {
 		return nil, err
@@ -6469,6 +6494,25 @@ func (s *SqlStore) IsClusterAddressConflicting(ctx context.Context, clusterAddre
 	return count > 0, nil
 }
 
+// HasForeignAccountProxyAtHost reports whether a proxy owned by a different
+// account declares this host. Shared proxies (account_id IS NULL) are not
+// foreign: a shared cluster is what most accounts pin their agent network
+// gateway to. The match folds case because proxies declare their address as
+// the operator spelled it while the caller's host is normalised; that costs a
+// scan of the proxies table, taken once per account when its gateway is
+// bootstrapped, not on the per-connect path IsClusterAddressConflicting serves.
+func (s *SqlStore) HasForeignAccountProxyAtHost(ctx context.Context, host, accountID string) (bool, error) {
+	var count int64
+	result := s.db.
+		Model(&proxy.Proxy{}).
+		Where("LOWER(cluster_address) = LOWER(?) AND account_id IS NOT NULL AND account_id != ?", host, accountID).
+		Count(&count)
+	if result.Error != nil {
+		return false, status.Errorf(status.Internal, "check proxy host ownership: %v", result.Error)
+	}
+	return count > 0, nil
+}
+
 func (s *SqlStore) DeleteAccountCluster(ctx context.Context, clusterAddress, accountID string) error {
 	result := s.db.
 		Where("cluster_address = ? AND account_id = ?", clusterAddress, accountID).

management/server/store/sql_store_agentnetwork.go

@@ -315,6 +315,36 @@ func (s *SqlStore) GetAllAgentNetworkSettings(ctx context.Context, lockStrength
 	return settings, nil
 }
 
+// HasGatewayClusterPinnedByOtherAccount reports whether another account has a
+// labeled agent network gateway pinned beneath host, making host its cluster.
+// A self-addressed endpoint on the very same hostname is not counted: that
+// collision is the domain unique index's to refuse, as a conflict. Case-folded,
+// since a settings row written before hostnames were normalised may carry
+// capitals; one row per account, so the scan is cheap.
+func (s *SqlStore) HasGatewayClusterPinnedByOtherAccount(ctx context.Context, host, accountID string) (bool, error) {
+	return s.countGatewayRowsByOtherAccount(ctx, "LOWER(proxy_address) = LOWER(?) AND LOWER(domain) <> LOWER(proxy_address)", host, accountID)
+}
+
+// HasGatewayEndpointByOtherAccount reports whether host is another account's
+// agent network endpoint hostname (domain). Case-folded for the same reason as
+// HasGatewayClusterPinnedByOtherAccount.
+func (s *SqlStore) HasGatewayEndpointByOtherAccount(ctx context.Context, host, accountID string) (bool, error) {
+	return s.countGatewayRowsByOtherAccount(ctx, "LOWER(domain) = LOWER(?)", host, accountID)
+}
+
+func (s *SqlStore) countGatewayRowsByOtherAccount(ctx context.Context, predicate, host, accountID string) (bool, error) {
+	var count int64
+	result := s.db.
+		Model(&agentNetworkTypes.Settings{}).
+		Where(predicate+" AND account_id != ?", host, accountID).
+		Count(&count)
+	if result.Error != nil {
+		log.WithContext(ctx).Errorf("failed to check agent network gateway claims at host: %v", result.Error)
+		return false, status.Errorf(status.Internal, "check agent network gateway claims at host")
+	}
+	return count > 0, nil
+}
+
 // GetAgentNetworkSettingsByProxyAddress returns every Settings row whose
 // gateway is served by the proxy declaring the given cluster address. Used by
 // cluster-scoped synthesis to find the accounts a shared proxy serves.

management/server/store/store.go

@@ -4,6 +4,7 @@ package store
 
 import (
 	"context"
+	"database/sql"
 	"errors"
 	"fmt"
 	"net"
@@ -15,6 +16,7 @@ import (
 	"runtime"
 	"slices"
 	"strings"
+	"sync"
 	"time"
 
 	"github.com/google/uuid"
@@ -340,6 +342,9 @@ type Store interface {
 	CountProxiesByAccountID(ctx context.Context, accountID string) (int64, error)
 	IsClusterAddressConflicting(ctx context.Context, clusterAddress, accountID string) (bool, error)
 	HasActiveProxyAtClusterAddress(ctx context.Context, clusterAddress string) (bool, error)
+	HasForeignAccountProxyAtHost(ctx context.Context, host, accountID string) (bool, error)
+	HasGatewayClusterPinnedByOtherAccount(ctx context.Context, host, accountID string) (bool, error)
+	HasGatewayEndpointByOtherAccount(ctx context.Context, host, accountID string) (bool, error)
 	DeleteAccountCluster(ctx context.Context, clusterAddress, accountID string) error
 
 	GetCustomDomainsCounts(ctx context.Context) (total int64, validated int64, err error)
@@ -732,6 +737,7 @@ func NewTestStoreFromSQL(ctx context.Context, filename string, dataDir string) (
 			time.Sleep(100 * time.Millisecond)
 		}
 	}
+	store.Close(ctx)
 	return nil, nil, fmt.Errorf("failed to create test store after %d attempts: %v", maxRetries, err)
 }
 
@@ -758,14 +764,15 @@ func addAllGroupToAccount(ctx context.Context, store Store) error {
 	return nil
 }
 
-func getSqlStoreEngine(ctx context.Context, store *SqlStore, kind types.Engine) (Store, func(), error) {
+func getSqlStoreEngine(ctx context.Context, sqliteStore *SqlStore, kind types.Engine) (Store, func(), error) {
+	store := sqliteStore
 	var cleanup func()
 	var err error
 	switch kind {
 	case types.PostgresStoreEngine:
-		store, cleanup, err = newReusedPostgresStore(ctx, store, kind)
+		store, cleanup, err = newReusedPostgresStore(ctx, sqliteStore, kind)
 	case types.MysqlStoreEngine:
-		store, cleanup, err = newReusedMysqlStore(ctx, store, kind)
+		store, cleanup, err = newReusedMysqlStore(ctx, sqliteStore, kind)
 	default:
 		cleanup = func() {
 			// sqlite doesn't need to be cleaned up
@@ -781,6 +788,11 @@ func getSqlStoreEngine(ctx context.Context, store *SqlStore, kind types.Engine)
 		if store.pool != nil {
 			store.pool.Close()
 		}
+		if store != sqliteStore {
+			// The sqlite store only seeded the engine under test; without this
+			// every test leaks its connection and the opener goroutines.
+			sqliteStore.Close(ctx)
+		}
 	}
 
 	return store, closeConnection, nil
@@ -805,19 +817,23 @@ func newReusedPostgresStore(ctx context.Context, store *SqlStore, kind types.Eng
 		return nil, nil, fmt.Errorf("failed to open postgres connection: %v", err)
 	}
 
-	dsn, cleanup, err := createRandomDB(dsn, db, kind)
-
-	sqlDB, _ := db.DB()
-	if sqlDB != nil {
-		sqlDB.Close()
+	template, err := postgresSchemaTemplate(ctx, dsn, db)
+	if err != nil {
+		closeGormDB(db)
+		return nil, nil, err
 	}
 
+	dsn, cleanup, err := createRandomDB(dsn, db, kind, template)
+
+	closeGormDB(db)
+
 	if err != nil {
 		return nil, nil, err
 	}
 
-	store, err = NewPostgresqlStoreFromSqlStore(ctx, store, dsn, nil)
+	store, err = newPostgresqlStoreFromSqlStore(ctx, store, dsn, nil, true)
 	if err != nil {
+		cleanup()
 		return nil, nil, err
 	}
 
@@ -850,7 +866,13 @@ func newReusedMysqlStore(ctx context.Context, store *SqlStore, kind types.Engine
 	sqlDB.SetMaxOpenConns(1)
 	sqlDB.SetMaxIdleConns(1)
 
-	dsn, cleanup, err := createRandomDB(dsn, db, kind)
+	tableDDL, err := mysqlSchemaTemplate(ctx, dsn, db)
+	if err != nil {
+		sqlDB.Close()
+		return nil, nil, err
+	}
+
+	dsn, cleanup, err := createRandomDB(dsn, db, kind, "")
 
 	sqlDB.Close()
 
@@ -858,14 +880,200 @@ func newReusedMysqlStore(ctx context.Context, store *SqlStore, kind types.Engine
 		return nil, nil, err
 	}
 
-	store, err = NewMysqlStoreFromSqlStore(ctx, store, dsn, nil)
+	if err := cloneMysqlSchema(ctx, dsn, tableDDL); err != nil {
+		cleanup()
+		return nil, nil, err
+	}
+
+	store, err = newMysqlStoreFromSqlStore(ctx, store, dsn, nil, true)
 	if err != nil {
+		cleanup()
 		return nil, nil, err
 	}
 
 	return store, cleanup, nil
 }
 
+// schemaTemplates remembers, per engine and server, a database that went
+// through the full migration once in this process. Every later test database
+// is cloned from it, so a test pays for CREATE DATABASE and a schema copy
+// instead of the 40-table AutoMigrate plus every pre and post migration, which
+// is what made each MySQL test store cost well over a second in CI.
+var (
+	schemaTemplatesMu sync.Mutex
+	schemaTemplates   = map[string]*schemaTemplate{}
+)
+
+type schemaTemplate struct {
+	dbName string
+	// tableDDL holds the CREATE TABLE statements of the template. MySQL has no
+	// server-side database template, so the schema is replayed statement by
+	// statement into each test database.
+	tableDDL []string
+}
+
+func schemaTemplateKey(engine types.Engine, dsn string) string {
+	return string(engine) + "|" + dsn
+}
+
+func newTestDBName(prefix string) string {
+	return fmt.Sprintf("%s_%s", prefix, strings.ReplaceAll(uuid.New().String(), "-", "_"))
+}
+
+// postgresSchemaTemplate returns the name of a fully migrated database that
+// CREATE DATABASE ... TEMPLATE can copy, creating it on first use.
+func postgresSchemaTemplate(ctx context.Context, baseDSN string, admin *gorm.DB) (string, error) {
+	schemaTemplatesMu.Lock()
+	defer schemaTemplatesMu.Unlock()
+
+	key := schemaTemplateKey(types.PostgresStoreEngine, baseDSN)
+	if tpl, ok := schemaTemplates[key]; ok {
+		return tpl.dbName, nil
+	}
+
+	name := newTestDBName("test_template")
+	if err := admin.Exec(fmt.Sprintf("CREATE DATABASE %s", name)).Error; err != nil {
+		return "", fmt.Errorf("create postgres template database: %w", err)
+	}
+
+	tplStore, err := NewPostgresqlStoreForTests(ctx, replaceDBName(baseDSN, name), nil, false)
+	if err != nil {
+		dropDatabase(admin, name)
+		return "", fmt.Errorf("migrate postgres template database: %w", err)
+	}
+	// TEMPLATE refuses a source that still has sessions, so release both handles
+	// before the first clone.
+	tplStore.Close(ctx)
+	if tplStore.pool != nil {
+		tplStore.pool.Close()
+	}
+
+	schemaTemplates[key] = &schemaTemplate{dbName: name}
+	return name, nil
+}
+
+// mysqlSchemaTemplate returns the CREATE TABLE statements of a fully migrated
+// database, migrating one on first use.
+func mysqlSchemaTemplate(ctx context.Context, baseDSN string, admin *gorm.DB) ([]string, error) {
+	schemaTemplatesMu.Lock()
+	defer schemaTemplatesMu.Unlock()
+
+	key := schemaTemplateKey(types.MysqlStoreEngine, baseDSN)
+	if tpl, ok := schemaTemplates[key]; ok {
+		return tpl.tableDDL, nil
+	}
+
+	name := newTestDBName("test_template")
+	if err := admin.Exec(fmt.Sprintf("CREATE DATABASE %s", name)).Error; err != nil {
+		return nil, fmt.Errorf("create mysql template database: %w", err)
+	}
+
+	tplStore, err := NewMysqlStore(ctx, replaceDBName(baseDSN, name), nil, false)
+	if err != nil {
+		dropDatabase(admin, name)
+		return nil, fmt.Errorf("migrate mysql template database: %w", err)
+	}
+	tableDDL, err := mysqlTableDDL(ctx, tplStore.db, name)
+	tplStore.Close(ctx)
+	if err != nil {
+		dropDatabase(admin, name)
+		return nil, err
+	}
+
+	schemaTemplates[key] = &schemaTemplate{dbName: name, tableDDL: tableDDL}
+	return tableDDL, nil
+}
+
+func mysqlTableDDL(ctx context.Context, db *gorm.DB, dbName string) ([]string, error) {
+	sqlDB, err := db.DB()
+	if err != nil {
+		return nil, err
+	}
+
+	tables, err := mysqlTableNames(ctx, sqlDB, dbName)
+	if err != nil {
+		return nil, err
+	}
+
+	tableDDL := make([]string, 0, len(tables))
+	for _, table := range tables {
+		var name, createStmt string
+		row := sqlDB.QueryRowContext(ctx, fmt.Sprintf("SHOW CREATE TABLE %s.%s", dbName, table))
+		if err := row.Scan(&name, &createStmt); err != nil {
+			return nil, fmt.Errorf("read create statement of %s: %w", table, err)
+		}
+		tableDDL = append(tableDDL, createStmt)
+	}
+	return tableDDL, nil
+}
+
+func mysqlTableNames(ctx context.Context, sqlDB *sql.DB, dbName string) ([]string, error) {
+	rows, err := sqlDB.QueryContext(ctx, fmt.Sprintf("SHOW TABLES FROM %s", dbName))
+	if err != nil {
+		return nil, fmt.Errorf("list template tables: %w", err)
+	}
+	defer rows.Close()
+
+	var tables []string
+	for rows.Next() {
+		var table string
+		if err := rows.Scan(&table); err != nil {
+			return nil, fmt.Errorf("scan template table name: %w", err)
+		}
+		tables = append(tables, table)
+	}
+	if err := rows.Err(); err != nil {
+		return nil, fmt.Errorf("list template tables: %w", err)
+	}
+	return tables, nil
+}
+
+// cloneMysqlSchema replays the template's CREATE TABLE statements into the
+// database the DSN points at.
+func cloneMysqlSchema(ctx context.Context, dsn string, tableDDL []string) error {
+	db, err := gorm.Open(mysql.Open(dsn+"?charset=utf8&parseTime=True&loc=Local"), getGormConfig())
+	if err != nil {
+		return fmt.Errorf("connect to test database: %w", err)
+	}
+	sqlDB, err := db.DB()
+	if err != nil {
+		return err
+	}
+	defer sqlDB.Close()
+
+	// The statements come out of SHOW TABLES in name order, not dependency
+	// order, and their foreign keys reference tables of the session's default
+	// database. Pin a single connection so the session setting below covers
+	// every statement, and connect straight to the new database so unqualified
+	// references land there.
+	sqlDB.SetMaxOpenConns(1)
+	if _, err := sqlDB.ExecContext(ctx, "SET FOREIGN_KEY_CHECKS = 0"); err != nil {
+		return fmt.Errorf("disable foreign key checks: %w", err)
+	}
+	for _, stmt := range tableDDL {
+		if _, err := sqlDB.ExecContext(ctx, stmt); err != nil {
+			return fmt.Errorf("replay table definition: %w", err)
+		}
+	}
+	return nil
+}
+
+// dropDatabase removes a template that never became usable, so a failed setup
+// does not leave it behind on a shared server. The server may still be tearing
+// down the sessions the failed migration held, so the drop retries while
+// Postgres reports the database as in use.
+func dropDatabase(admin *gorm.DB, name string) {
+	if err := execWithTemplateRetry(admin, fmt.Sprintf("DROP DATABASE IF EXISTS %s", name)); err != nil {
+		log.Warnf("failed to drop template database %s: %v", name, err)
+	}
+}
+
+func closeGormDB(db *gorm.DB) {
+	if sqlDB, _ := db.DB(); sqlDB != nil {
+		sqlDB.Close()
+	}
+}
+
 func openDBWithRetry(dsn string, engine types.Engine, maxRetries int) (*gorm.DB, error) {
 	var db *gorm.DB
 	var err error
@@ -891,10 +1099,16 @@ func openDBWithRetry(dsn string, engine types.Engine, maxRetries int) (*gorm.DB,
 	return nil, err
 }
 
-func createRandomDB(dsn string, db *gorm.DB, engine types.Engine) (string, func(), error) {
-	dbName := fmt.Sprintf("test_db_%s", strings.ReplaceAll(uuid.New().String(), "-", "_"))
+// createRandomDB creates a uniquely named database for one test. On postgres a
+// non-empty template is copied server-side with CREATE DATABASE ... TEMPLATE.
+func createRandomDB(dsn string, db *gorm.DB, engine types.Engine, template string) (string, func(), error) {
+	dbName := newTestDBName("test_db")
 
-	if err := db.Exec(fmt.Sprintf("CREATE DATABASE %s", dbName)).Error; err != nil {
+	createStmt := fmt.Sprintf("CREATE DATABASE %s", dbName)
+	if template != "" && engine == types.PostgresStoreEngine {
+		createStmt = fmt.Sprintf("CREATE DATABASE %s TEMPLATE %s", dbName, template)
+	}
+	if err := execWithTemplateRetry(db, createStmt); err != nil {
 		return "", nil, fmt.Errorf("failed to create database: %v", err)
 	}
 
@@ -960,6 +1174,20 @@ func createRandomDB(dsn string, db *gorm.DB, engine types.Engine) (string, func(
 	return replaceDBName(dsn, dbName), cleanup, nil
 }
 
+// execWithTemplateRetry runs a statement, retrying briefly when postgres still
+// sees the template's just-closed sessions and refuses to copy it.
+func execWithTemplateRetry(db *gorm.DB, stmt string) error {
+	var err error
+	for attempt := 0; attempt < 20; attempt++ {
+		err = db.Exec(stmt).Error
+		if err == nil || !strings.Contains(err.Error(), "is being accessed by other users") {
+			return err
+		}
+		time.Sleep(100 * time.Millisecond)
+	}
+	return err
+}
+
 func replaceDBName(dsn, newDBName string) string {
 	re := regexp.MustCompile(`(?P<pre>[:/@])(?P<dbname>[^/?]+)(?P<post>\?|$)`)
 	return re.ReplaceAllString(dsn, `${pre}`+newDBName+`${post}`)

management/server/store/store_mock.go

@@ -3065,6 +3065,51 @@ func (mr *MockStoreMockRecorder) HasActiveProxyAtClusterAddress(ctx, clusterAddr
 	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "HasActiveProxyAtClusterAddress", reflect.TypeOf((*MockStore)(nil).HasActiveProxyAtClusterAddress), ctx, clusterAddress)
 }
 
+// HasForeignAccountProxyAtHost mocks base method.
+func (m *MockStore) HasForeignAccountProxyAtHost(ctx context.Context, host, accountID string) (bool, error) {
+	m.ctrl.T.Helper()
+	ret := m.ctrl.Call(m, "HasForeignAccountProxyAtHost", ctx, host, accountID)
+	ret0, _ := ret[0].(bool)
+	ret1, _ := ret[1].(error)
+	return ret0, ret1
+}
+
+// HasForeignAccountProxyAtHost indicates an expected call of HasForeignAccountProxyAtHost.
+func (mr *MockStoreMockRecorder) HasForeignAccountProxyAtHost(ctx, host, accountID any) *gomock.Call {
+	mr.mock.ctrl.T.Helper()
+	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "HasForeignAccountProxyAtHost", reflect.TypeOf((*MockStore)(nil).HasForeignAccountProxyAtHost), ctx, host, accountID)
+}
+
+// HasGatewayClusterPinnedByOtherAccount mocks base method.
+func (m *MockStore) HasGatewayClusterPinnedByOtherAccount(ctx context.Context, host, accountID string) (bool, error) {
+	m.ctrl.T.Helper()
+	ret := m.ctrl.Call(m, "HasGatewayClusterPinnedByOtherAccount", ctx, host, accountID)
+	ret0, _ := ret[0].(bool)
+	ret1, _ := ret[1].(error)
+	return ret0, ret1
+}
+
+// HasGatewayClusterPinnedByOtherAccount indicates an expected call of HasGatewayClusterPinnedByOtherAccount.
+func (mr *MockStoreMockRecorder) HasGatewayClusterPinnedByOtherAccount(ctx, host, accountID any) *gomock.Call {
+	mr.mock.ctrl.T.Helper()
+	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "HasGatewayClusterPinnedByOtherAccount", reflect.TypeOf((*MockStore)(nil).HasGatewayClusterPinnedByOtherAccount), ctx, host, accountID)
+}
+
+// HasGatewayEndpointByOtherAccount mocks base method.
+func (m *MockStore) HasGatewayEndpointByOtherAccount(ctx context.Context, host, accountID string) (bool, error) {
+	m.ctrl.T.Helper()
+	ret := m.ctrl.Call(m, "HasGatewayEndpointByOtherAccount", ctx, host, accountID)
+	ret0, _ := ret[0].(bool)
+	ret1, _ := ret[1].(error)
+	return ret0, ret1
+}
+
+// HasGatewayEndpointByOtherAccount indicates an expected call of HasGatewayEndpointByOtherAccount.
+func (mr *MockStoreMockRecorder) HasGatewayEndpointByOtherAccount(ctx, host, accountID any) *gomock.Call {
+	mr.mock.ctrl.T.Helper()
+	return mr.mock.ctrl.RecordCallWithMethodType(mr.mock, "HasGatewayEndpointByOtherAccount", reflect.TypeOf((*MockStore)(nil).HasGatewayEndpointByOtherAccount), ctx, host, accountID)
+}
+
 // IncrementAgentNetworkConsumption mocks base method.
 func (m *MockStore) IncrementAgentNetworkConsumption(ctx context.Context, accountID string, kind types.ConsumptionDimension, dimID string, windowSeconds int64, windowStart time.Time, tokensIn, tokensOut int64, costUSD float64) error {
 	m.ctrl.T.Helper()

management/server/testutil/store.go

@@ -37,6 +37,18 @@ func CreateMysqlTestContainer() (func(), string, error) {
 		mysql.WithDatabase("testing"),
 		mysql.WithUsername("root"),
 		mysql.WithPassword("testing"),
+		// Every test creates and drops a database with about 40 tables, so with
+		// the server defaults the run is dominated by durability work: each
+		// CREATE TABLE fsyncs the redo log, the binary log and the doublewrite
+		// buffer. None of it protects anything in a container that is discarded
+		// after the run. Tables stay in per-table files on purpose: in the shared
+		// system tablespace the cost of every CREATE and DROP grew with the number
+		// of databases the run had already created.
+		testcontainers.WithCmd("mysqld",
+			"--innodb-flush-log-at-trx-commit=0",
+			"--innodb-doublewrite=OFF",
+			"--skip-log-bin",
+		),
 		testcontainers.WithWaitStrategy(
 			wait.ForLog("/usr/sbin/mysqld: ready for connections").
 				WithOccurrence(1).WithStartupTimeout(15*time.Second).WithPollInterval(100*time.Millisecond),

management/server/types/user.go

@@ -285,6 +285,25 @@ func (u *User) EncryptSensitiveData(enc *crypt.FieldEncrypt) error {
 	return nil
 }
 
+func MaskEmail(email string) string {
+	local, domain, found := strings.Cut(email, "@")
+	if !found || local == "" || domain == "" {
+		return ""
+	}
+
+	// Runes, not bytes, so a non-ASCII local part is not cut mid-character.
+	runes := []rune(local)
+
+	// Keeping the first two and the last needs a local part of at least four to
+	// hide anything at all: at three or fewer those are the whole of it, and the
+	// address would be recoverable in full from what is meant to conceal it.
+	if len(runes) < 4 {
+		return "****@" + domain
+	}
+
+	return string(runes[:2]) + "****" + string(runes[len(runes)-1]) + "@" + domain
+}
+
 // DecryptSensitiveData decrypts the user's sensitive fields (Email and Name) in place.
 func (u *User) DecryptSensitiveData(enc *crypt.FieldEncrypt) error {
 	if enc == nil {

management/server/types/user_test.go

@@ -296,3 +296,144 @@ func TestUser_EncryptDecryptRoundTrip(t *testing.T) {
 		})
 	}
 }
+
+func TestMaskEmail(t *testing.T) {
+	testCases := []struct {
+		name     string
+		email    string
+		expected string
+	}{
+		{
+			name:     "ordinary address keeps the first two, the last, and the domain",
+			email:    "admin@example.com",
+			expected: "ad****n@example.com",
+		},
+		{
+			name:     "four characters is the shortest local part that reveals anything",
+			email:    "abcd@example.com",
+			expected: "ab****d@example.com",
+		},
+		{
+			name:     "three character local part is masked whole, since a lead and tail would be all of it",
+			email:    "abc@example.com",
+			expected: "****@example.com",
+		},
+		{
+			name:     "two character local part is masked whole",
+			email:    "ab@example.com",
+			expected: "****@example.com",
+		},
+		{
+			name:     "single character local part is masked whole",
+			email:    "a@b.co",
+			expected: "****@b.co",
+		},
+		{
+			name:     "mask width does not report the length it stands in for",
+			email:    "a.very.long.local.part@example.com",
+			expected: "a.****t@example.com",
+		},
+		{
+			name:     "a local part far longer than the mask is still reduced to three characters",
+			email:    "finance.department.notifications.owner.account@example.com",
+			expected: "fi****t@example.com",
+		},
+		{
+			name:     "plus addressing is masked along with the rest of the local part",
+			email:    "admin+netbird@example.com",
+			expected: "ad****d@example.com",
+		},
+		{
+			name:     "separators inside the local part are not treated specially",
+			email:    "first.last-name_x@example.com",
+			expected: "fi****x@example.com",
+		},
+		{
+			name:     "case is preserved rather than normalised",
+			email:    "Admin@Example.COM",
+			expected: "Ad****n@Example.COM",
+		},
+		{
+			name:     "subdomains stay intact",
+			email:    "owner@mail.corp.example.com",
+			expected: "ow****r@mail.corp.example.com",
+		},
+		{
+			name:     "german umlauts count as single characters",
+			email:    "müller@example.de",
+			expected: "mü****r@example.de",
+		},
+		{
+			name:     "cyrillic local part is cut on runes",
+			email:    "иванов@example.ru",
+			expected: "ив****в@example.ru",
+		},
+		{
+			name:     "cjk local part of three runes is masked whole, counted in runes not bytes",
+			email:    "用户名@example.cn",
+			expected: "****@example.cn",
+		},
+		{
+			name:     "cjk local part of four runes reveals the first two and the last",
+			email:    "用户名字@example.cn",
+			expected: "用户****字@example.cn",
+		},
+		{
+			name:     "arabic local part is cut on runes",
+			email:    "مستخدم@example.sa",
+			expected: "مس****م@example.sa",
+		},
+		{
+			name:     "two rune non-ascii local part is masked whole",
+			email:    "ää@example.de",
+			expected: "****@example.de",
+		},
+		{
+			name:     "astral plane runes are not split into surrogates",
+			email:    "a🎉bc@example.com",
+			expected: "a🎉****c@example.com",
+		},
+		{
+			name:     "a non-ascii domain is left alone",
+			email:    "admin@münchen.example",
+			expected: "ad****n@münchen.example",
+		},
+		{
+			name:     "only the first separator splits, so a second stays in the domain",
+			email:    "a@b@example.com",
+			expected: "****@b@example.com",
+		},
+		{
+			name:     "empty email has nothing to mask",
+			email:    "",
+			expected: "",
+		},
+		{
+			name:     "value without a separator is not an address",
+			email:    "not-an-email",
+			expected: "",
+		},
+		{
+			name:     "missing local part is not an address",
+			email:    "@example.com",
+			expected: "",
+		},
+		{
+			name:     "missing domain is not an address",
+			email:    "admin@",
+			expected: "",
+		},
+		{
+			name:     "a bare separator is not an address",
+			email:    "@",
+			expected: "",
+		},
+	}
+
+	for _, tc := range testCases {
+		t.Run(tc.name, func(t *testing.T) {
+			assert.Equal(t, tc.expected, MaskEmail(tc.email))
+		})
+	}
+
+}

management/server/user.go

@@ -1373,6 +1373,25 @@ func (am *DefaultAccountManager) deleteRegularUser(ctx context.Context, accountI
 	return updateAccountPeers, nil
 }
 
+// pendingApprovalError refuses a user awaiting approval, naming the owner who
+// can approve them when their address resolves. Failing to resolve one is not a
+// reason to withhold the refusal, so the lookup is best effort.
+func (am *DefaultAccountManager) pendingApprovalError(ctx context.Context, accountID string) error {
+	owner, err := am.GetOwnerInfo(ctx, accountID)
+	if err != nil {
+		log.WithContext(ctx).Debugf("pending approval refusal: owner of account %s did not resolve: %v", accountID, err)
+		return status.NewUserPendingApprovalError()
+	}
+
+	masked := types.MaskEmail(owner.Email)
+	if masked == "" {
+		log.WithContext(ctx).Debugf("pending approval refusal: no address found for the owner of account %s", accountID)
+		return status.NewUserPendingApprovalError()
+	}
+
+	return status.NewUserPendingApprovalByOwnerError(masked)
+}
+
 // GetOwnerInfo retrieves the owner information for a given account ID.
 func (am *DefaultAccountManager) GetOwnerInfo(ctx context.Context, accountID string) (*types.UserInfo, error) {
 	owner, err := am.Store.GetAccountOwner(ctx, store.LockingStrengthNone, accountID)
@@ -1430,6 +1449,14 @@ func (am *DefaultAccountManager) GetCurrentUserInfo(ctx context.Context, userAut
 		return nil, err
 	}
 
+	// A user pending approval is blocked too, and the dashboard needs to tell
+	// the two apart: one is a dead end, the other resolves by itself once the
+	// owner acts. Naming that owner needs the address the IdP holds, which is
+	// why this is answered here rather than in the permission gate.
+	if user.IsBlocked() && user.PendingApproval {
+		return nil, am.pendingApprovalError(ctx, user.AccountID)
+	}
+
 	if user.IsBlocked() {
 		return nil, status.NewUserBlockedError()
 	}

management/server/user_test.go

@@ -1826,6 +1826,42 @@ func TestDefaultAccountManager_GetCurrentUserInfo(t *testing.T) {
 	}
 	require.NoError(t, store.SaveAccount(context.Background(), account2))
 
+	account3 := newAccountWithId(context.Background(), "account3", "account3Owner", "", "owner@example.com", "", false)
+	account3.Users["pending-user"] = &types.User{
+		Id:              "pending-user",
+		AccountID:       account3.Id,
+		Role:            types.UserRoleUser,
+		Blocked:         true,
+		PendingApproval: true,
+	}
+	require.NoError(t, store.SaveAccount(context.Background(), account3))
+
+	// The owner has no address to name, so the refusal falls back to the generic one.
+	account4 := newAccountWithId(context.Background(), "account4", "account4Owner", "", "", "", false)
+	account4.Users["pending-user-without-owner-email"] = &types.User{
+		Id:              "pending-user-without-owner-email",
+		AccountID:       account4.Id,
+		Role:            types.UserRoleUser,
+		Blocked:         true,
+		PendingApproval: true,
+	}
+	require.NoError(t, store.SaveAccount(context.Background(), account4))
+
+	// No user holds the owner role, so the owner lookup itself fails.
+	account5 := newAccountWithId(context.Background(), "account5", "account5Admin", "", "", "", false)
+	account5.Users["account5Admin"].Role = types.UserRoleAdmin
+	account5.Users["pending-user-without-owner"] = &types.User{
+		Id:              "pending-user-without-owner",
+		AccountID:       account5.Id,
+		Role:            types.UserRoleUser,
+		Blocked:         true,
+		PendingApproval: true,
+	}
+	require.NoError(t, store.SaveAccount(context.Background(), account5))
+
+	account6 := newAccountWithId(context.Background(), "account6", "account6Owner", "", "stranger@example.com", "", false)
+	require.NoError(t, store.SaveAccount(context.Background(), account6))
+
 	permissionsManager := permissions.NewManager(store)
 	am := DefaultAccountManager{
 		Store:              store,
@@ -1854,6 +1890,34 @@ func TestDefaultAccountManager_GetCurrentUserInfo(t *testing.T) {
 			userAuth:    auth.UserAuth{AccountId: account1.Id, UserId: "service-user"},
 			expectedErr: status.NewPermissionDeniedError(),
 		},
+		{
+			name:        "pending approval names the owner",
+			userAuth:    auth.UserAuth{AccountId: account3.Id, UserId: "pending-user"},
+			expectedErr: status.NewUserPendingApprovalByOwnerError("ow****r@example.com"),
+		},
+		{
+			name:        "pending approval without an owner address",
+			userAuth:    auth.UserAuth{AccountId: account4.Id, UserId: "pending-user-without-owner-email"},
+			expectedErr: status.NewUserPendingApprovalError(),
+		},
+		{
+			name:        "pending approval without an owner",
+			userAuth:    auth.UserAuth{AccountId: account5.Id, UserId: "pending-user-without-owner"},
+			expectedErr: status.NewUserPendingApprovalError(),
+		},
+		{
+			// The account claim points at an account the caller is not in. The
+			// owner named has to be the one of the account holding the caller's
+			// own record, never the one the claim asks for.
+			name:        "pending approval ignores a mismatched account claim",
+			userAuth:    auth.UserAuth{AccountId: account6.Id, UserId: "pending-user"},
+			expectedErr: status.NewUserPendingApprovalByOwnerError("ow****r@example.com"),
+		},
+		{
+			name:        "blocked user answers before the account claim is validated",
+			userAuth:    auth.UserAuth{AccountId: account6.Id, UserId: "blocked-user"},
+			expectedErr: status.NewUserBlockedError(),
+		},
 		{
 			name:     "owner user",
 			userAuth: auth.UserAuth{AccountId: account1.Id, UserId: "account1Owner"},