0.10.1: 安全与业务逻辑加固、新品牌与部署加固

三轮审查修复(60+ 项),相对远端 main(b536672)的关键变更: - 安全: 数据面 SSRF 拨号防护(防 DNS rebinding)/上游凭据剥离/登录防枚举与锁定态统一/可信代理(X-Forwarded-For)限流加固/会话版本失效机制/ 撤销即时传播/弱密钥拒绝启动/脱敏字节级重写(保签名契约) - 业务逻辑: 裸 body 上传 panic/bootstrap 审计管线卡死/定价通配符优先级/ 全局工具可见性/调度器停机补跑/TOTP 挑战令牌消费顺序/熔断探针语义/ >4MB 响应 token 计量/管理员重置密码作废会话等 - 前端: 新 logo(语枢 AI 网关主题)/Provider 凭据异常警示/删除入口/ 后端错误消息透传/localStorage 敏感数据收敛 - 部署: CREDENTIAL_MASTER_KEY 持久化与弱值拒绝/Provider DELETE 接口/ nginx 安全头/worker 内存限制 - 新增迁移 000029(key_hash 索引)/000030(usage_daily 币种维度)
2026-08-13 10:50:51 +08:00
parent b536672000
commit 9501751792
136 changed files with 8024 additions and 1476 deletions
@@ -0,0 +1,168 @@
+package scheduler
+
+import (
+	"context"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"testing"
+	"time"
+
+	"aigateway.local/core/internal/platform/config"
+	"aigateway.local/core/internal/platform/cryptox"
+	"aigateway.local/core/internal/platform/database"
+)
+
+func TestSchedulerPostgreSQLLifecycle(t *testing.T) {
+	databaseURL := os.Getenv("SCHEDULER_TEST_DATABASE_URL")
+	if databaseURL == "" {
+		t.Skip("SCHEDULER_TEST_DATABASE_URL is not set")
+	}
+	ctx := context.Background()
+	pool, err := database.Open(ctx, config.Database{URL: databaseURL, MaxConns: 8})
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer pool.Close()
+	adminID := "64444444-4444-4444-8444-444444444444"
+	appID := "65555555-5555-4555-8555-555555555555"
+	versionID := "66666666-6666-4666-8666-666666666666"
+	cleanup := func() {
+		_, _ = pool.Exec(ctx, `DELETE FROM gateway.scheduled_tasks WHERE code='scheduler_test_task'`)
+		_, _ = pool.Exec(ctx, `DELETE FROM gateway.applications WHERE id=$1`, appID)
+		_, _ = pool.Exec(ctx, `DELETE FROM gateway.admin_accounts WHERE id=$1`, adminID)
+	}
+	cleanup()
+	defer cleanup()
+	if _, err = pool.Exec(ctx, `INSERT INTO gateway.admin_accounts(id,username,password_hash,role,active) VALUES($1,'scheduler-test-admin','test','superadmin',true)`, adminID); err != nil {
+		t.Fatal(err)
+	}
+	tx, err := pool.Begin(ctx)
+	if err != nil {
+		t.Fatal(err)
+	}
+	configJSON := `{"model":"test-model","knowledge_base_ids":[],"tool_ids":[],"retrieval_top_k":4,"temperature":0.2,"max_tool_rounds":1}`
+	if _, err = tx.Exec(ctx, `INSERT INTO gateway.applications(id,code,name,status,draft_config,created_by) VALUES($1,'scheduler_test_app','Scheduler Test App','active',$2,$3)`, appID, configJSON, adminID); err != nil {
+		t.Fatal(err)
+	}
+	if _, err = tx.Exec(ctx, `INSERT INTO gateway.application_versions(id,application_id,version,config,published_by) VALUES($1,$2,1,$3,$4)`, versionID, appID, configJSON, adminID); err != nil {
+		t.Fatal(err)
+	}
+	if _, err = tx.Exec(ctx, `UPDATE gateway.applications SET published_version=1 WHERE id=$1`, appID); err != nil {
+		t.Fatal(err)
+	}
+	if err = tx.Commit(ctx); err != nil {
+		t.Fatal(err)
+	}
+
+	requestCount := 0
+	failRequests := false
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		requestCount++
+		if r.URL.Path != "/v1/applications/scheduler_test_app/chat/completions" || r.Header.Get("X-Gateway-API-Key") != "gw_scheduler_test" {
+			http.Error(w, "unexpected request", http.StatusUnauthorized)
+			return
+		}
+		if failRequests {
+			http.Error(w, `{"error":"temporary failure"}`, http.StatusServiceUnavailable)
+			return
+		}
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{"choices":[{"message":{"role":"assistant","content":"scheduled answer"}}]}`))
+	}))
+	defer server.Close()
+	cipher, err := cryptox.NewKeyring("AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA=", 1, "", "scheduled-task-api-key")
+	if err != nil {
+		t.Fatal(err)
+	}
+	service := NewService(pool, cipher)
+	task, err := service.Save(ctx, "", TaskInput{Code: "scheduler_test_task", Name: "Scheduler Test", CronExpression: "*/5 * * * *", Timezone: "UTC", TargetType: "application", TargetCode: "scheduler_test_app", Prompt: "create report", Variables: json.RawMessage(`{"scope":"daily"}`), APIKey: "gw_scheduler_test"}, adminID)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !task.HasAPIKey || task.Enabled {
+		t.Fatalf("unexpected task: %+v", task)
+	}
+	if _, err = service.QueueManual(ctx, task.ID); err != nil {
+		t.Fatal(err)
+	}
+	engine := NewEngine(service, server.URL, "integration-worker", 10, 3, time.Minute, nil)
+	processed, err := engine.Tick(ctx, time.Now())
+	if err != nil || processed != 1 || requestCount != 1 {
+		t.Fatalf("tick processed=%d requests=%d err=%v", processed, requestCount, err)
+	}
+	runs, err := service.Runs(ctx, task.ID, 10)
+	if err != nil || len(runs) != 1 || runs[0].Status != "success" || responseAnswer(runs[0].Response) != "scheduled answer" {
+		t.Fatalf("runs=%+v err=%v", runs, err)
+	}
+	var completed bool
+	if err = pool.QueryRow(ctx, `SELECT EXISTS(SELECT 1 FROM gateway.outbox_events WHERE aggregate_type='scheduled_task' AND aggregate_id=$1 AND event_type='scheduled_task.completed')`, runs[0].ID).Scan(&completed); err != nil || !completed {
+		t.Fatalf("completion event=%v err=%v", completed, err)
+	}
+
+	// Force a due schedule in the past. Tick must enqueue it once, execute it,
+	// and advance next_run_at beyond now rather than replay every missed slot.
+	if _, err = service.SetEnabled(ctx, task.ID, true); err != nil {
+		t.Fatal(err)
+	}
+	forcedNow := time.Now().UTC()
+	if _, err = pool.Exec(ctx, `UPDATE gateway.scheduled_tasks SET next_run_at=$2 WHERE id=$1`, task.ID, forcedNow.Add(-time.Minute)); err != nil {
+		t.Fatal(err)
+	}
+	processed, err = engine.Tick(ctx, forcedNow)
+	if err != nil || processed != 1 || requestCount != 2 {
+		t.Fatalf("due tick processed=%d requests=%d err=%v", processed, requestCount, err)
+	}
+	task, err = service.Get(ctx, task.ID)
+	if err != nil || task.NextRunAt == nil || !task.NextRunAt.After(forcedNow) {
+		t.Fatalf("next run was not advanced: %+v err=%v", task.NextRunAt, err)
+	}
+
+	// Ordinary gateway failures remain pending until the configured attempt
+	// limit, then become failed and emit exactly one terminal event.
+	if _, err = service.SetEnabled(ctx, task.ID, false); err != nil {
+		t.Fatal(err)
+	}
+	failRequests = true
+	failedRun, err := service.QueueManual(ctx, task.ID)
+	if err != nil {
+		t.Fatal(err)
+	}
+	for attempt := 1; attempt <= 3; attempt++ {
+		processed, err = engine.Tick(ctx, time.Now())
+		if err != nil || processed != 1 {
+			t.Fatalf("retry tick attempt=%d processed=%d err=%v", attempt, processed, err)
+		}
+		current, getErr := service.getRun(ctx, failedRun.ID)
+		wantStatus := "pending"
+		if attempt == 3 {
+			wantStatus = "failed"
+		}
+		if getErr != nil || current.Status != wantStatus || current.Attempts != attempt {
+			t.Fatalf("retry attempt=%d run=%+v err=%v", attempt, current, getErr)
+		}
+	}
+	var failedEvents int
+	if err = pool.QueryRow(ctx, `SELECT count(*) FROM gateway.outbox_events WHERE aggregate_type='scheduled_task' AND aggregate_id=$1 AND event_type='scheduled_task.failed'`, failedRun.ID).Scan(&failedEvents); err != nil || failedEvents != 1 {
+		t.Fatalf("failure events=%d err=%v", failedEvents, err)
+	}
+
+	// A worker lease that stays running past its timeout is finalized through
+	// the same failed-run and outbox path once it reaches the attempt limit.
+	staleRun, err := service.QueueManual(ctx, task.ID)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if _, err = pool.Exec(ctx, `UPDATE gateway.scheduled_task_runs SET status='running',attempts=3,worker_id='dead-worker',started_at=$2 WHERE id=$1`, staleRun.ID, time.Now().Add(-2*time.Minute)); err != nil {
+		t.Fatal(err)
+	}
+	processed, err = engine.Tick(ctx, time.Now())
+	if err != nil || processed != 0 {
+		t.Fatalf("stale tick processed=%d err=%v", processed, err)
+	}
+	staleRun, err = service.getRun(ctx, staleRun.ID)
+	if err != nil || staleRun.Status != "failed" || staleRun.Error != "执行超时且达到最大重试次数" {
+		t.Fatalf("stale run=%+v err=%v", staleRun, err)
+	}
+}