Browse Source

fix(node-sync): keep shared-email lookups on the open transaction

emailUsedByOtherInbounds and emailsUsedByOtherInbounds always queried
through database.GetDB(), yet five callers run them inside a transaction:
both node traffic sync sweeps (setRemoteTrafficLocked), the inbound edit
stats sync (updateClientTraffics) and both email-drop branches of
UpdateInboundClient. Each call needed a second pool connection while the
transaction held row locks on client_traffics/node_client_traffics. On
PostgreSQL, once subscription fetches (UPDATE client_traffics SET
last_sub_fetch) blocked on those locks filled the rest of the 25-connection
pool, the lookup waited forever, the transaction never ended, and the panel
hung until restart.

Both helpers now take the *gorm.DB to query on; transactional callers pass
tx and the rest pass the pool handle they already hold. The new tests cap
the pool at one connection, so any query that escapes the transaction
blocks; each went red before the fix on SQLite and (node sync) on a live
PostgreSQL 16, and passes after.

Closes #6789
MHSanaei 21 hours ago
parent
commit
b5a04ea95f

+ 1 - 1
internal/web/service/client_bulk.go

@@ -1143,7 +1143,7 @@ func (s *ClientService) bulkDelInboundClients(
 	var sharedSet map[string]bool
 	if !keepTraffic {
 		var sharedErr error
-		sharedSet, sharedErr = inboundSvc.emailsUsedByOtherInbounds(foundList, inboundId)
+		sharedSet, sharedErr = inboundSvc.emailsUsedByOtherInbounds(db, foundList, inboundId)
 		if sharedErr != nil {
 			for email := range wantedEmails {
 				res.perEmailSkipped[email] = sharedErr.Error()

+ 2 - 2
internal/web/service/client_identity_normalized_test.go

@@ -92,14 +92,14 @@ func TestEmailsUsedByOtherInboundsMatchesCaseInsensitively(t *testing.T) {
 		}
 	}
 
-	shared, err := is.emailsUsedByOtherInbounds([]string{"alice@x"}, ibA.Id)
+	shared, err := is.emailsUsedByOtherInbounds(database.GetDB(), []string{"alice@x"}, ibA.Id)
 	if err != nil {
 		t.Fatalf("emailsUsedByOtherInbounds: %v", err)
 	}
 	if !shared["alice@x"] {
 		t.Error("lower-cased lookup missed a stored mixed-case email")
 	}
-	used, err := is.emailUsedByOtherInbounds("alice@x", ibA.Id)
+	used, err := is.emailUsedByOtherInbounds(database.GetDB(), "alice@x", ibA.Id)
 	if err != nil {
 		t.Fatalf("emailUsedByOtherInbounds: %v", err)
 	}

+ 4 - 4
internal/web/service/client_inbound_apply.go

@@ -136,7 +136,7 @@ func (s *ClientService) delInboundClients(inboundSvc *InboundService, inboundId
 			}
 		}
 		var sharedErr error
-		sharedSet, sharedErr = inboundSvc.emailsUsedByOtherInbounds(removedEmails, inboundId)
+		sharedSet, sharedErr = inboundSvc.emailsUsedByOtherInbounds(db, removedEmails, inboundId)
 		if sharedErr != nil {
 			return false, sharedErr
 		}
@@ -946,7 +946,7 @@ func (s *ClientService) UpdateInboundClient(inboundSvc *InboundService, data *mo
 						return e
 					}
 				} else {
-					stillUsed, sErr := inboundSvc.emailUsedByOtherInbounds(oldEmail, data.Id)
+					stillUsed, sErr := inboundSvc.emailUsedByOtherInbounds(tx, oldEmail, data.Id)
 					if sErr != nil {
 						return sErr
 					}
@@ -968,7 +968,7 @@ func (s *ClientService) UpdateInboundClient(inboundSvc *InboundService, data *mo
 				}
 			}
 		} else {
-			stillUsed, sErr := inboundSvc.emailUsedByOtherInbounds(oldEmail, data.Id)
+			stillUsed, sErr := inboundSvc.emailUsedByOtherInbounds(tx, oldEmail, data.Id)
 			if sErr != nil {
 				return sErr
 			}
@@ -1140,7 +1140,7 @@ func (s *ClientService) DelInboundClientByEmail(inboundSvc *InboundService, inbo
 	prevSettings := oldInbound.Settings
 	oldInbound.Settings = string(newSettings)
 
-	emailShared, err := inboundSvc.emailUsedByOtherInbounds(email, inboundId)
+	emailShared, err := inboundSvc.emailUsedByOtherInbounds(db, email, inboundId)
 	if err != nil {
 		return false, err
 	}

+ 1 - 1
internal/web/service/inbound.go

@@ -2290,7 +2290,7 @@ func (s *InboundService) updateClientTraffics(tx *gorm.DB, oldInbound *model.Inb
 		if _, kept := newEmails[email]; kept {
 			continue
 		}
-		stillUsed, err := s.emailUsedByOtherInbounds(email, oldInbound.Id)
+		stillUsed, err := s.emailUsedByOtherInbounds(tx, email, oldInbound.Id)
 		if err != nil {
 			return err
 		}

+ 5 - 6
internal/web/service/inbound_clients.go

@@ -124,14 +124,14 @@ func (s *InboundService) backfillClientStats(db *gorm.DB, inbounds []*model.Inbo
 	return clientsByInbound
 }
 
-// emailUsedByOtherInbounds reports whether email lives in any inbound other
-// than exceptInboundId. Empty email returns false.
-func (s *InboundService) emailUsedByOtherInbounds(email string, exceptInboundId int) (bool, error) {
+// emailUsedByOtherInbounds reports whether email lives in any inbound other than
+// exceptInboundId. Inside a transaction pass tx: a pool query there starves the pool.
+func (s *InboundService) emailUsedByOtherInbounds(db *gorm.DB, email string, exceptInboundId int) (bool, error) {
 	if email == "" {
 		return false, nil
 	}
 	var count int64
-	err := database.GetDB().Table("client_inbounds").
+	err := db.Table("client_inbounds").
 		Joins("JOIN clients ON clients.id = client_inbounds.client_id").
 		Where("client_inbounds.inbound_id != ? AND LOWER(clients.email) = ?",
 			exceptInboundId, strings.ToLower(strings.TrimSpace(email))).
@@ -142,7 +142,7 @@ func (s *InboundService) emailUsedByOtherInbounds(email string, exceptInboundId
 	return count > 0, nil
 }
 
-func (s *InboundService) emailsUsedByOtherInbounds(emails []string, exceptInboundId int) (map[string]bool, error) {
+func (s *InboundService) emailsUsedByOtherInbounds(db *gorm.DB, emails []string, exceptInboundId int) (map[string]bool, error) {
 	shared := make(map[string]bool, len(emails))
 	want := make(map[string]struct{}, len(emails))
 	for _, e := range emails {
@@ -158,7 +158,6 @@ func (s *InboundService) emailsUsedByOtherInbounds(emails []string, exceptInboun
 	for e := range want {
 		lowered = append(lowered, e)
 	}
-	db := database.GetDB()
 	for _, batch := range chunkStrings(lowered, sqlInChunk) {
 		var rows []struct{ Email string }
 		err := db.Table("client_inbounds").

+ 2 - 2
internal/web/service/inbound_node.go

@@ -940,7 +940,7 @@ func (s *InboundService) setRemoteTrafficLocked(nodeID int, snap *runtime.Traffi
 			// email's last inbound — wiping it while a sibling still feeds it
 			// loses the summed history, and the next node sync would re-seed the
 			// row with that node's counter alone.
-			sharedEmails, sErr := s.emailsUsedByOtherInbounds(goneEmails, c.Id)
+			sharedEmails, sErr := s.emailsUsedByOtherInbounds(tx, goneEmails, c.Id)
 			if sErr != nil {
 				return false, sErr
 			}
@@ -1218,7 +1218,7 @@ func (s *InboundService) setRemoteTrafficLocked(nodeID int, snap *runtime.Traffi
 			}
 			// Same shared-accumulator rule as the inbound-removal sweep above:
 			// keep the row while another inbound still references the email.
-			stillUsed, uErr := s.emailUsedByOtherInbounds(existing.Email, c.Id)
+			stillUsed, uErr := s.emailUsedByOtherInbounds(tx, existing.Email, c.Id)
 			if uErr != nil {
 				return false, uErr
 			}

+ 180 - 0
internal/web/service/tx_pool_starvation_test.go

@@ -0,0 +1,180 @@
+package service
+
+import (
+	"testing"
+	"time"
+
+	"gorm.io/gorm"
+
+	"github.com/mhsanaei/3x-ui/v3/internal/database"
+	"github.com/mhsanaei/3x-ui/v3/internal/database/model"
+	"github.com/mhsanaei/3x-ui/v3/internal/web/runtime"
+	"github.com/mhsanaei/3x-ui/v3/internal/xray"
+)
+
+// capPoolAtOneConn models a pool whose other connections are all parked on the
+// transaction's row locks: any query outside tx then waits forever.
+func capPoolAtOneConn(t *testing.T) {
+	t.Helper()
+	sqlDB, err := database.GetDB().DB()
+	if err != nil {
+		t.Fatalf("sql.DB: %v", err)
+	}
+	sqlDB.SetMaxOpenConns(1)
+}
+
+func runWithoutSecondConn(t *testing.T, name string, fn func() error) error {
+	t.Helper()
+	done := make(chan error, 1)
+	go func() { done <- fn() }()
+	select {
+	case err := <-done:
+		return err
+	case <-time.After(5 * time.Second):
+		// The DB cleanup's Close fails the parked pool request, which unwinds fn.
+		t.Fatalf("%s blocked waiting for a second pool connection while its transaction is open", name)
+		return nil
+	}
+}
+
+func createNodeInboundWithSettings(t *testing.T, db *gorm.DB, nodeID int, tag string, port int, settings string) {
+	t.Helper()
+	nid := nodeID
+	ib := &model.Inbound{UserId: 1, Tag: tag, Enable: true, Port: port, Protocol: model.VLESS, NodeID: &nid, Settings: settings}
+	if err := db.Create(ib).Error; err != nil {
+		t.Fatalf("create node inbound %q: %v", tag, err)
+	}
+}
+
+func syncNodeSnapshot(t *testing.T, svc *InboundService, nodeID int, inbounds ...*model.Inbound) error {
+	t.Helper()
+	_, err := svc.setRemoteTrafficLocked(nodeID, &runtime.TrafficSnapshot{Inbounds: inbounds}, false, false)
+	return err
+}
+
+func assertNoTraffic(t *testing.T, email string) {
+	t.Helper()
+	var n int64
+	if err := database.GetDB().Model(xray.ClientTraffic{}).Where("email = ?", email).Count(&n).Error; err != nil {
+		t.Fatalf("count traffic %q: %v", email, err)
+	}
+	if n != 0 {
+		t.Fatalf("traffic rows for %q = %d, want 0", email, n)
+	}
+}
+
+func TestNodeSyncInboundRemovalStaysOnOneConnection(t *testing.T) {
+	db := initTrafficTestDB(t)
+	goneSettings := `{"clients": [{"email": "gone", "enable": true}]}`
+	keepSettings := `{"clients": [{"email": "keep", "enable": true}]}`
+	createNodeInboundWithSettings(t, db, 1, "n1-gone", 41001, goneSettings)
+	createNodeInboundWithSettings(t, db, 1, "n1-keep", 41002, keepSettings)
+	svc := &InboundService{}
+	keep := &model.Inbound{Tag: "n1-keep", Enable: true, Settings: keepSettings, ClientStats: []xray.ClientTraffic{{Email: "keep", Up: 10, Down: 10, Enable: true}}}
+	if err := syncNodeSnapshot(t, svc, 1,
+		&model.Inbound{Tag: "n1-gone", Enable: true, Settings: goneSettings, ClientStats: []xray.ClientTraffic{{Email: "gone", Up: 10, Down: 10, Enable: true}}},
+		keep); err != nil {
+		t.Fatalf("seed sync: %v", err)
+	}
+	readTraffic(t, db, "gone")
+
+	capPoolAtOneConn(t)
+	err := runWithoutSecondConn(t, "node sync inbound-removal sweep", func() error {
+		return syncNodeSnapshot(t, svc, 1, keep)
+	})
+	if err != nil {
+		t.Fatalf("sync node 1 without n1-gone: %v", err)
+	}
+	assertNoTraffic(t, "gone")
+}
+
+func TestNodeSyncClientGoneStaysOnOneConnection(t *testing.T) {
+	db := initTrafficTestDB(t)
+	bothSettings := `{"clients": [{"email": "gone", "enable": true}, {"email": "keep", "enable": true}]}`
+	createNodeInboundWithSettings(t, db, 1, "n1-in", 41001, bothSettings)
+	svc := &InboundService{}
+	if err := syncNodeSnapshot(t, svc, 1, &model.Inbound{Tag: "n1-in", Enable: true, Settings: bothSettings, ClientStats: []xray.ClientTraffic{
+		{Email: "gone", Up: 10, Down: 10, Enable: true},
+		{Email: "keep", Up: 10, Down: 10, Enable: true},
+	}}); err != nil {
+		t.Fatalf("seed sync: %v", err)
+	}
+	readTraffic(t, db, "gone")
+
+	capPoolAtOneConn(t)
+	err := runWithoutSecondConn(t, "node sync per-client sweep", func() error {
+		return syncNodeSnapshot(t, svc, 1, &model.Inbound{
+			Tag:         "n1-in",
+			Enable:      true,
+			Settings:    `{"clients": [{"email": "keep", "enable": true}]}`,
+			ClientStats: []xray.ClientTraffic{{Email: "keep", Up: 20, Down: 20, Enable: true}},
+		})
+	})
+	if err != nil {
+		t.Fatalf("sync node 1 without the gone client: %v", err)
+	}
+	var baselines int64
+	if err := db.Model(&model.NodeClientTraffic{}).Where("node_id = ? AND email = ?", 1, "gone").Count(&baselines).Error; err != nil {
+		t.Fatalf("count node baselines: %v", err)
+	}
+	if baselines != 0 {
+		t.Fatalf("node baseline for the gone client = %d rows, want 0", baselines)
+	}
+}
+
+func TestUpdateClientTrafficsStaysOnOneConnection(t *testing.T) {
+	setupBulkDB(t)
+	svc := &InboundService{}
+	clients := []model.Client{{Email: "gone@x", ID: "aaaaaaaa-0000-0000-0000-000000000101", Enable: true}}
+	oldIb := mkInbound(t, 22101, model.VLESS, clientsSettings(t, clients))
+	if err := database.GetDB().Create(&xray.ClientTraffic{InboundId: oldIb.Id, Email: "gone@x", Enable: true}).Error; err != nil {
+		t.Fatalf("seed traffic: %v", err)
+	}
+	newIb := *oldIb
+	newIb.Settings = `{"clients": []}`
+
+	capPoolAtOneConn(t)
+	err := runWithoutSecondConn(t, "inbound edit dropping a client", func() error {
+		return database.GetDB().Transaction(func(tx *gorm.DB) error {
+			return svc.updateClientTraffics(tx, oldIb, &newIb)
+		})
+	})
+	if err != nil {
+		t.Fatalf("updateClientTraffics: %v", err)
+	}
+	var n int64
+	database.GetDB().Model(xray.ClientTraffic{}).Where("email = ?", "gone@x").Count(&n)
+	if n != 0 {
+		t.Fatalf("traffic rows for dropped client = %d, want 0", n)
+	}
+}
+
+func TestUpdateInboundClientRenameOntoExistingStatStaysOnOneConnection(t *testing.T) {
+	setupBulkDB(t)
+	svc := &ClientService{}
+	inboundSvc := &InboundService{}
+	source := []model.Client{{Email: "old@x", ID: "aaaaaaaa-0000-0000-0000-000000000102", SubID: "sub-102", Enable: true}}
+	ib := mkInbound(t, 22102, model.VLESS, clientsSettings(t, source))
+	if err := svc.SyncInbound(nil, ib.Id, source); err != nil {
+		t.Fatalf("seed linkage: %v", err)
+	}
+	if err := database.GetDB().Create(&xray.ClientTraffic{InboundId: ib.Id, Email: "new@x", Enable: true}).Error; err != nil {
+		t.Fatalf("seed target traffic: %v", err)
+	}
+
+	renamed := source
+	renamed[0].Email = "new@x"
+	capPoolAtOneConn(t)
+	err := runWithoutSecondConn(t, "client rename onto an existing stats row", func() error {
+		_, err := svc.UpdateInboundClient(inboundSvc, &model.Inbound{Id: ib.Id, Settings: clientsSettings(t, renamed)}, "old@x")
+		return err
+	})
+	if err != nil {
+		t.Fatalf("UpdateInboundClient: %v", err)
+	}
+	var n int64
+	database.GetDB().Model(xray.ClientTraffic{}).Where("email = ?", "old@x").Count(&n)
+	if n != 0 {
+		t.Fatalf("traffic rows for renamed-away email = %d, want 0", n)
+	}
+}