fix(backend): resolve SQLite deadlock in tunnel entry updates (#334)

- Fix deadlock when updating tunnel entries with offline nodes
- Add test file for tunnel entry SQLite operations
- Update contract tests for entry port conflict and limiter sync
- Add plan documents for SQLite deadlock fix and contract semantics
This commit is contained in:
sagit
2026-03-18 09:00:20 +08:00
committed by GitHub
parent 41ef814643
commit 2df061a19f
7 changed files with 253 additions and 17 deletions
+21 -7
View File
@@ -832,7 +832,7 @@ func (h *Handler) tunnelUpdate(w http.ResponseWriter, r *http.Request) {
newEntryNodeIDs = append(newEntryNodeIDs, in.NodeID) newEntryNodeIDs = append(newEntryNodeIDs, in.NodeID)
} }
} }
if err := h.validateTunnelEntryPortConflictsForNewEntries(id, oldEntryNodeIDs, newEntryNodeIDs); err != nil { if err := h.validateTunnelEntryPortConflictsForNewEntries(tx, id, oldEntryNodeIDs, newEntryNodeIDs); err != nil {
response.WriteJSON(w, response.ErrDefault(err.Error())) response.WriteJSON(w, response.ErrDefault(err.Error()))
return return
} }
@@ -1020,8 +1020,8 @@ func (h *Handler) cleanupTunnelForwardRuntimesOnRemovedEntryNodes(tunnelID int64
} }
} }
func (h *Handler) validateTunnelEntryPortConflictsForNewEntries(tunnelID int64, oldEntryNodeIDs, newEntryNodeIDs []int64) error { func (h *Handler) validateTunnelEntryPortConflictsForNewEntries(tx *gorm.DB, tunnelID int64, oldEntryNodeIDs, newEntryNodeIDs []int64) error {
if h == nil || h.repo == nil || tunnelID <= 0 { if h == nil || h.repo == nil || tx == nil || tunnelID <= 0 {
return nil return nil
} }
@@ -1030,7 +1030,7 @@ func (h *Handler) validateTunnelEntryPortConflictsForNewEntries(tunnelID int64,
return nil return nil
} }
forwards, err := h.listForwardsByTunnel(tunnelID) forwards, err := h.repo.ListForwardsByTunnelTx(tx, tunnelID)
if err != nil || len(forwards) == 0 { if err != nil || len(forwards) == 0 {
return nil return nil
} }
@@ -1040,7 +1040,7 @@ func (h *Handler) validateTunnelEntryPortConflictsForNewEntries(tunnelID int64,
if f == nil { if f == nil {
continue continue
} }
oldPorts, portsErr := h.listForwardPorts(f.ID) oldPorts, portsErr := h.repo.ListForwardPortsTx(tx, f.ID)
if portsErr != nil { if portsErr != nil {
continue continue
} }
@@ -1050,14 +1050,14 @@ func (h *Handler) validateTunnelEntryPortConflictsForNewEntries(tunnelID int64,
} }
for _, nodeID := range addedNodeIDs { for _, nodeID := range addedNodeIDs {
node, nodeErr := h.getNodeRecord(nodeID) node, nodeErr := h.repo.GetNodeRecordTx(tx, nodeID)
if nodeErr != nil { if nodeErr != nil {
continue continue
} }
if err := validateLocalNodePort(node, port); err != nil { if err := validateLocalNodePort(node, port); err != nil {
return fmt.Errorf("转发 %s 入口端口冲突: %w", f.Name, err) return fmt.Errorf("转发 %s 入口端口冲突: %w", f.Name, err)
} }
if err := h.validateForwardPortAvailability(node, port, f.ID); err != nil { if err := h.validateForwardPortAvailabilityTx(tx, node, port, f.ID); err != nil {
return fmt.Errorf("转发 %s 入口端口冲突: %w", f.Name, err) return fmt.Errorf("转发 %s 入口端口冲突: %w", f.Name, err)
} }
} }
@@ -4153,6 +4153,20 @@ func (h *Handler) validateForwardPortAvailability(node *nodeRecord, port int, cu
return nil return nil
} }
func (h *Handler) validateForwardPortAvailabilityTx(tx *gorm.DB, node *nodeRecord, port int, currentForwardID int64) error {
if h == nil || h.repo == nil || tx == nil || node == nil || port <= 0 {
return nil
}
occupied, err := h.repo.HasOtherForwardOnNodePortTx(tx, node.ID, port, currentForwardID)
if err != nil {
return err
}
if occupied {
return fmt.Errorf("节点 %s 端口 %d 已被其他转发占用", node.Name, port)
}
return nil
}
func parsePortRangeMinMax(input string) (int, int) { func parsePortRangeMinMax(input string) (int, int) {
input = strings.TrimSpace(input) input = strings.TrimSpace(input)
if input == "" { if input == "" {
@@ -0,0 +1,104 @@
package handler
import (
"path/filepath"
"testing"
"time"
"go-backend/internal/store/repo"
)
func TestValidateTunnelEntryPortConflictsForNewEntriesDoesNotBlockOnSQLiteTx(t *testing.T) {
r, err := repo.Open(filepath.Join(t.TempDir(), "panel.db"))
if err != nil {
t.Fatalf("open sqlite: %v", err)
}
t.Cleanup(func() {
_ = r.Close()
})
h := &Handler{repo: r}
now := time.Now().UnixMilli()
if err := r.DB().Exec(`
INSERT INTO node(name, secret, server_ip, port, created_time, status, tcp_listen_addr, udp_listen_addr, is_remote)
VALUES
('entry-old', 'secret-old', '10.0.0.1', '12000-12010', ?, 1, '[::]', '[::]', 0),
('entry-new', 'secret-new', '10.0.0.2', '12000-12010', ?, 1, '[::]', '[::]', 0)
`, now, now).Error; err != nil {
t.Fatalf("insert nodes: %v", err)
}
var oldEntryID, newEntryID int64
if err := r.DB().Raw(`SELECT id FROM node WHERE name = 'entry-old'`).Scan(&oldEntryID).Error; err != nil {
t.Fatalf("load old entry id: %v", err)
}
if err := r.DB().Raw(`SELECT id FROM node WHERE name = 'entry-new'`).Scan(&newEntryID).Error; err != nil {
t.Fatalf("load new entry id: %v", err)
}
if err := r.DB().Exec(`
INSERT INTO tunnel(name, traffic_ratio, type, protocol, flow, created_time, updated_time, status, inx, ip_preference)
VALUES('sqlite-tunnel', 1, 1, 'tls', 1, ?, ?, 1, 1, '')
`, now, now).Error; err != nil {
t.Fatalf("insert tunnel: %v", err)
}
var tunnelID int64
if err := r.DB().Raw(`SELECT id FROM tunnel WHERE name = 'sqlite-tunnel'`).Scan(&tunnelID).Error; err != nil {
t.Fatalf("load tunnel id: %v", err)
}
if err := r.DB().Exec(`
INSERT INTO chain_tunnel(tunnel_id, chain_type, node_id, inx, protocol)
VALUES(?, '1', ?, 1, 'tls')
`, tunnelID, oldEntryID).Error; err != nil {
t.Fatalf("insert chain_tunnel: %v", err)
}
if err := r.DB().Exec(`
INSERT INTO forward(user_id, user_name, name, tunnel_id, remote_addr, strategy, created_time, updated_time, status, inx)
VALUES(1, 'tester', 'forward-a', ?, '127.0.0.1:8080', 'fifo', ?, ?, 1, 1)
`, tunnelID, now, now).Error; err != nil {
t.Fatalf("insert forward: %v", err)
}
var forwardID int64
if err := r.DB().Raw(`SELECT id FROM forward WHERE name = 'forward-a'`).Scan(&forwardID).Error; err != nil {
t.Fatalf("load forward id: %v", err)
}
if err := r.DB().Exec(`
INSERT INTO forward_port(forward_id, node_id, port)
VALUES(?, ?, 12001)
`, forwardID, oldEntryID).Error; err != nil {
t.Fatalf("insert forward_port: %v", err)
}
tx := r.BeginTx()
if tx == nil {
t.Fatal("begin tx: nil transaction")
}
if tx.Error != nil {
t.Fatalf("begin tx: %v", tx.Error)
}
errCh := make(chan error, 1)
doneCh := make(chan struct{})
go func() {
defer close(doneCh)
errCh <- h.validateTunnelEntryPortConflictsForNewEntries(tx, tunnelID, []int64{oldEntryID}, []int64{oldEntryID, newEntryID})
}()
select {
case err := <-errCh:
if err != nil {
_ = tx.Rollback().Error
t.Fatalf("unexpected validation error: %v", err)
}
case <-time.After(500 * time.Millisecond):
_ = tx.Rollback().Error
<-doneCh
t.Fatal("validation blocked while transaction was open on sqlite")
}
if err := tx.Rollback().Error; err != nil {
t.Fatalf("rollback tx: %v", err)
}
}
@@ -30,8 +30,15 @@ func (r *Repository) ListForwardsByTunnel(tunnelID int64) ([]model.ForwardRecord
if r == nil || r.db == nil { if r == nil || r.db == nil {
return nil, errors.New("repository not initialized") return nil, errors.New("repository not initialized")
} }
return r.ListForwardsByTunnelTx(r.db, tunnelID)
}
func (r *Repository) ListForwardsByTunnelTx(tx *gorm.DB, tunnelID int64) ([]model.ForwardRecord, error) {
if tx == nil {
return nil, errors.New("database unavailable")
}
var forwards []model.Forward var forwards []model.Forward
err := r.db.Where("tunnel_id = ?", tunnelID).Order("id ASC").Find(&forwards).Error err := tx.Where("tunnel_id = ?", tunnelID).Order("id ASC").Find(&forwards).Error
if err != nil { if err != nil {
return nil, err return nil, err
} }
@@ -95,8 +102,15 @@ func (r *Repository) ListForwardPorts(forwardID int64) ([]model.ForwardPortRecor
if r == nil || r.db == nil { if r == nil || r.db == nil {
return nil, errors.New("repository not initialized") return nil, errors.New("repository not initialized")
} }
return r.ListForwardPortsTx(r.db, forwardID)
}
func (r *Repository) ListForwardPortsTx(tx *gorm.DB, forwardID int64) ([]model.ForwardPortRecord, error) {
if tx == nil {
return nil, errors.New("database unavailable")
}
var ports []model.ForwardPort var ports []model.ForwardPort
err := r.db.Where("forward_id = ?", forwardID).Order("id ASC").Find(&ports).Error err := tx.Where("forward_id = ?", forwardID).Order("id ASC").Find(&ports).Error
if err != nil { if err != nil {
return nil, err return nil, err
} }
@@ -115,12 +129,19 @@ func (r *Repository) HasOtherForwardOnNodePort(nodeID int64, port int, currentFo
if r == nil || r.db == nil { if r == nil || r.db == nil {
return false, errors.New("repository not initialized") return false, errors.New("repository not initialized")
} }
return r.HasOtherForwardOnNodePortTx(r.db, nodeID, port, currentForwardID)
}
func (r *Repository) HasOtherForwardOnNodePortTx(tx *gorm.DB, nodeID int64, port int, currentForwardID int64) (bool, error) {
if tx == nil {
return false, errors.New("database unavailable")
}
if nodeID <= 0 || port <= 0 { if nodeID <= 0 || port <= 0 {
return false, nil return false, nil
} }
var count int64 var count int64
err := r.db.Model(&model.ForwardPort{}). err := tx.Model(&model.ForwardPort{}).
Where("node_id = ? AND port = ? AND forward_id <> ?", nodeID, port, currentForwardID). Where("node_id = ? AND port = ? AND forward_id <> ?", nodeID, port, currentForwardID).
Count(&count).Error Count(&count).Error
if err != nil { if err != nil {
@@ -32,7 +32,6 @@ func TestIssue313_EntryPortCrossTunnelConflictContract(t *testing.T) {
return mustLastInsertID(t, repo, name) return mustLastInsertID(t, repo, name)
} }
entryA := insertNode("issue313-entry-a", "10.100.0.1", "2000-2010")
entryB1 := insertNode("issue313-entry-b1", "10.100.0.2", "2000-2010") entryB1 := insertNode("issue313-entry-b1", "10.100.0.2", "2000-2010")
entryB2 := insertNode("issue313-entry-b2", "10.100.0.3", "2000-2010") entryB2 := insertNode("issue313-entry-b2", "10.100.0.3", "2000-2010")
chainA := insertNode("issue313-chain-a", "10.100.0.4", "3000-3010") chainA := insertNode("issue313-chain-a", "10.100.0.4", "3000-3010")
@@ -51,7 +50,7 @@ func TestIssue313_EntryPortCrossTunnelConflictContract(t *testing.T) {
if err := repo.DB().Exec(` if err := repo.DB().Exec(`
INSERT INTO chain_tunnel(tunnel_id, chain_type, node_id, port, strategy, inx, protocol) INSERT INTO chain_tunnel(tunnel_id, chain_type, node_id, port, strategy, inx, protocol)
VALUES(?, 1, ?, 2000, 'round', 1, 'tls') VALUES(?, 1, ?, 2000, 'round', 1, 'tls')
`, tunnelAID, entryA).Error; err != nil { `, tunnelAID, entryB2).Error; err != nil {
t.Fatalf("insert chain_tunnel entry a: %v", err) t.Fatalf("insert chain_tunnel entry a: %v", err)
} }
if err := repo.DB().Exec(` if err := repo.DB().Exec(`
@@ -109,7 +108,7 @@ func TestIssue313_EntryPortCrossTunnelConflictContract(t *testing.T) {
} }
forwardAID := mustLastInsertID(t, repo, "issue313-forward-a") forwardAID := mustLastInsertID(t, repo, "issue313-forward-a")
if err := repo.DB().Exec(`INSERT INTO forward_port(forward_id, node_id, port) VALUES(?, ?, ?)`, forwardAID, entryA, 2000).Error; err != nil { if err := repo.DB().Exec(`INSERT INTO forward_port(forward_id, node_id, port) VALUES(?, ?, ?)`, forwardAID, entryB2, 2000).Error; err != nil {
t.Fatalf("insert forward_port a: %v", err) t.Fatalf("insert forward_port a: %v", err)
} }
@@ -171,7 +170,7 @@ func TestIssue313_EntryPortCrossTunnelConflictContract(t *testing.T) {
t.Fatalf("expected update failure due to cross-tunnel port conflict, got success with code 0") t.Fatalf("expected update failure due to cross-tunnel port conflict, got success with code 0")
} }
if !bytes.Contains(res.Body.Bytes(), []byte("端口")) && !bytes.Contains(res.Body.Bytes(), []byte("占用")) { if !bytes.Contains([]byte(out.Msg), []byte("端口")) && !bytes.Contains([]byte(out.Msg), []byte("占用")) {
t.Fatalf("expected port conflict error message, got %q", out.Msg) t.Fatalf("expected port conflict error message, got %q", out.Msg)
} }
@@ -705,7 +705,7 @@ func TestTunnelUpdateChangesEntryNodeButLeavesOldForwardRuntimeContract(t *testi
} }
oldEntryNodeID := insertNode("issue281-old-entry", "issue281-old-entry-secret", "10.51.0.1", "51000-51010", 0) oldEntryNodeID := insertNode("issue281-old-entry", "issue281-old-entry-secret", "10.51.0.1", "51000-51010", 0)
newEntryNodeID := insertNode("issue281-new-entry", "issue281-new-entry-secret", "10.51.0.2", "52000-52010", 1) newEntryNodeID := insertNode("issue281-new-entry", "issue281-new-entry-secret", "10.51.0.2", "51000-51010", 1)
exitNodeID := insertNode("issue281-exit", "issue281-exit-secret", "10.51.0.3", "53000-53010", 2) exitNodeID := insertNode("issue281-exit", "issue281-exit-secret", "10.51.0.3", "53000-53010", 2)
if err := r.DB().Exec(` if err := r.DB().Exec(`
@@ -881,8 +881,8 @@ func TestTunnelUpdateEntryTransitionsCleanupForwardRuntimeContract(t *testing.T)
} }
entryA := insertNode("issue281-transition-entry-a", "issue281-transition-entry-a-secret", "10.52.0.1", "54000-54010", 0) entryA := insertNode("issue281-transition-entry-a", "issue281-transition-entry-a-secret", "10.52.0.1", "54000-54010", 0)
entryB := insertNode("issue281-transition-entry-b", "issue281-transition-entry-b-secret", "10.52.0.2", "55000-55010", 1) entryB := insertNode("issue281-transition-entry-b", "issue281-transition-entry-b-secret", "10.52.0.2", "54000-54010", 1)
entryC := insertNode("issue281-transition-entry-c", "issue281-transition-entry-c-secret", "10.52.0.3", "56000-56010", 2) entryC := insertNode("issue281-transition-entry-c", "issue281-transition-entry-c-secret", "10.52.0.3", "54000-54010", 2)
exitNodeID := insertNode("issue281-transition-exit", "issue281-transition-exit-secret", "10.52.0.4", "57000-57010", 3) exitNodeID := insertNode("issue281-transition-exit", "issue281-transition-exit-secret", "10.52.0.4", "57000-57010", 3)
if err := r.DB().Exec(` if err := r.DB().Exec(`
@@ -0,0 +1,49 @@
# 042 - SQLite 隧道编辑添加入口节点卡死排查
## Issue
- 现象:SQLite 数据库下,编辑已有隧道并新增入口节点时接口卡住;PostgreSQL 下同样操作正常。
- 初步判断:`tunnelUpdate` 在事务尚未提交时触发了额外 repository 读查询,SQLite 配置 `MaxOpenConns(1)`,容易在同一请求内形成自锁等待。
## Goal
- 找出 SQLite 与 PostgreSQL 行为差异的根因。
- 修复隧道编辑新增入口节点时的阻塞问题,同时不破坏现有的入口端口冲突校验。
- 补充最小回归测试,锁定“事务内校验不可再次占用根连接”的场景。
## Checklist
- [x] 复核 `tunnelUpdate` 在新增入口节点路径上的调用链,确认事务内哪些查询绕过了 `tx`。
- [x] 为相关 repository 查询补齐 `Tx` 版本,避免 SQLite 单连接下的自锁等待。
- [x] 调整 handler 中入口端口冲突校验,保证事务内全程复用同一个 `tx`。
- [x] 增加针对 SQLite 的回归测试,验证事务内校验不会阻塞。
- [x] 运行相关 handler/backend 测试并记录结果。
## Notes
- 重点关注 `validateTunnelEntryPortConflictsForNewEntries`:当前它在 `tx.Commit()` 前执行,但内部调用 `ListForwardsByTunnel` / `ListForwardPorts` / `HasOtherForwardOnNodePort` 等非事务查询。
- SQLite 在 `go-backend/internal/store/repo/repository.go` 中显式设置了 `SetMaxOpenConns(1)`,因此这种模式在 SQLite 下会比 PostgreSQL 更容易表现为“卡死”。
## 实际改动
- `go-backend/internal/store/repo/repository_control.go`
- 新增 `ListForwardsByTunnelTx`、`ListForwardPortsTx`、`HasOtherForwardOnNodePortTx`,并让原有非事务方法复用统一实现。
- `go-backend/internal/http/handler/mutations.go`
- `tunnelUpdate` 在事务内执行入口端口冲突校验时显式传入当前 `tx`。
- `validateTunnelEntryPortConflictsForNewEntries` 改为全程使用事务查询。
- 新增 `validateForwardPortAvailabilityTx`,避免事务内回落到根连接查询。
- `go-backend/internal/http/handler/tunnel_entry_sqlite_test.go`
- 新增 SQLite 回归测试,验证开启事务后执行新增入口校验不会阻塞。
## 测试结果
### 通过
```bash
cd go-backend && go test ./internal/http/handler/...
```
- 结果:通过。
### 额外检查
```bash
cd go-backend && go test ./tests/contract/...
```
- 结果:未全绿;当前失败集中在既有的入口端口语义合同用例:
- `TestIssue313_EntryPortCrossTunnelConflictContract`
- `TestTunnelUpdateChangesEntryNodeButLeavesOldForwardRuntimeContract`
- `TestTunnelUpdateEntryTransitionsCleanupForwardRuntimeContract`
- 备注:这些失败反映的是“新增/切换入口时端口校验预期”与现有合同用例之间的行为差异,不是本次 SQLite 事务自锁修复本身的编译或阻塞问题。
@@ -0,0 +1,49 @@
# 043 - Entry Transition Contract Semantics Alignment
## Issue
- SQLite 阻塞修复完成后,`go test ./tests/contract/...` 暴露出 3 个入口变更相关合同用例失败。
- 失败原因分成两类:
- Issue 313 用例的数据构造没有真正制造“新增入口节点已被其他转发占用”的冲突。
- Issue 281 回归用例在后续引入“入口端口必须落在节点端口范围内”后,仍沿用旧的非重叠端口范围数据,和当前产品语义不一致。
## Goal
- 对齐这 3 个合同测试与当前后端语义。
- 保持 SQLite 自锁修复不回退。
- 让入口节点切换/新增相关合同测试重新稳定通过。
## Checklist
- [x] 复核 3 个失败用例的测试数据与当前后端校验语义差异。
- [x] 调整 Issue 313 用例,确保新增入口节点确实命中“其他转发已占用同节点同端口”。
- [x] 调整 Issue 281 两个回归用例,使入口切换场景使用与保留端口兼容的节点端口范围。
- [x] 运行相关合同测试与 handler 测试并记录结果。
## Notes
- `validateTunnelEntryPortConflictsForNewEntries` 的占用判定复用 `HasOtherForwardOnNodePort`,语义是“同节点 + 同端口 + 其他转发”,不是全局无节点维度的端口唯一性。
- 入口切换测试当前更关注 `forward_port` 重建和旧节点运行时清理,因此测试数据应避免被端口范围校验提前拦截。
## 实际调整
- `go-backend/tests/contract/issue313_entry_port_conflict_contract_test.go`
- 将隧道 A 的入口节点改为复用即将添加到隧道 B 的 `entryB2`,确保新增入口时真正命中“同节点同端口已被其他转发占用”。
- 修正错误消息断言,直接检查 `out.Msg`,避免 JSON 解码后再读 `res.Body` 导致误判。
- `go-backend/tests/contract/limiter_sync_failure_contract_test.go`
- 将 issue 281 的新入口节点端口范围调整为包含原有保留端口,保持测试关注点在运行时清理/同步,而不是被后续引入的端口范围校验拦截。
## 测试结果
### 定向回归
```bash
cd go-backend && go test ./tests/contract/... -run 'TestIssue313_EntryPortCrossTunnelConflictContract|TestTunnelUpdateChangesEntryNodeButLeavesOldForwardRuntimeContract|TestTunnelUpdateEntryTransitionsCleanupForwardRuntimeContract' -v
```
- 结果:3/3 通过。
### Handler
```bash
cd go-backend && go test ./internal/http/handler/...
```
- 结果:通过。
### 全量合同测试
```bash
cd go-backend && go test ./tests/contract/...
```
- 结果:通过。