Files
sundynix-agentix/sundynix-shared/bus/bus_e2e_test.go
T
Blizzard e3b0a4f81c perf(dispatcher): 任务并发消费 —— 待审/慢任务不再阻塞其他任务
bus.ConsumeTasks 由单条串行改为限并发分发:每个任务进独立 worker goroutine,
并发上限 = 信号量 + 消费者 MaxAckPending(DISPATCHER_CONCURRENCY,默认 8)。
- 背压:并发满则在 select{sem, ctx.Done} 处等空位;关停时留消息不 ack 待重投。
- 健壮:worker 内 recover panic → Term(避免崩溃循环);ack/nak/span 收口在 worker。
- 并发安全已核:CircuitBreaker 有锁、Orchestrator.turns 有 turnMu、pool RWMutex、
  evaluate 本就异步。

效果:一个 HITL 待审任务(Handle 阻塞至多 5min)或长 LLM 生成不再冻结后续任务。
验证:单测 TestConcurrentConsume(A 阻塞时 B 完成);live 实测 HITL 停 waiting 期间
普通任务 3s 跑完且 HITL 不受影响。全模块 build+vet+test 全绿。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-24 15:03:00 +08:00

257 lines
7.0 KiB
Go
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
package bus_test
import (
"context"
"encoding/json"
"testing"
"time"
natsserver "github.com/nats-io/nats-server/v2/server"
natstest "github.com/nats-io/nats-server/v2/test"
"github.com/sundynix/sundynix-shared/bus"
"github.com/sundynix/sundynix-shared/contract"
)
// startEmbeddedNATS 启动一个内嵌、开启 JetStream 的 NATS 服务器,免 Docker。
func startEmbeddedNATS(t *testing.T) string {
t.Helper()
opts := natstest.DefaultTestOptions
opts.Port = -1 // 随机端口
opts.JetStream = true
opts.StoreDir = t.TempDir()
srv := natstest.RunServer(&opts)
if !srv.ReadyForConnections(5 * time.Second) {
t.Fatal("embedded NATS not ready")
}
t.Cleanup(srv.Shutdown)
_ = natsserver.Server{} // 触发包引用
return srv.ClientURL()
}
// TestTaskRoundTrip 模拟 Gateway 发布 → NATS → Dispatcher 消费 的完整任务流。
func TestTaskRoundTrip(t *testing.T) {
url := startEmbeddedNATS(t)
// --- Gateway 侧:连接并声明任务流 ---
gw, err := bus.Connect(url)
if err != nil {
t.Fatalf("gateway connect: %v", err)
}
defer gw.Close()
ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
defer cancel()
if err := gw.EnsureTaskStream(ctx); err != nil {
t.Fatalf("ensure stream: %v", err)
}
// --- Dispatcher 侧:连接并开始消费 ---
dp, err := bus.Connect(url)
if err != nil {
t.Fatalf("dispatcher connect: %v", err)
}
defer dp.Close()
got := make(chan *contract.Task, 1)
stop, err := dp.ConsumeTasks(ctx, func(_ context.Context, task *contract.Task) error {
got <- task
return nil
})
if err != nil {
t.Fatalf("consume: %v", err)
}
defer stop()
// --- Gateway 发布一个任务 ---
want := &contract.Task{
ID: "task_demo_001",
Graph: json.RawMessage(`{"nodes":[{"id":"n1","type":"agent"}],"edges":[]}`),
Meta: map[string]any{"user": "wt"},
}
seq, err := gw.PublishTask(ctx, want)
if err != nil {
t.Fatalf("publish: %v", err)
}
if seq == 0 {
t.Fatal("expected non-zero stream sequence")
}
// --- 断言 Dispatcher 收到同一个任务 ---
select {
case task := <-got:
if task.ID != want.ID {
t.Fatalf("task id = %q, want %q", task.ID, want.ID)
}
if task.Meta["user"] != "wt" {
t.Fatalf("task meta lost: %+v", task.Meta)
}
t.Logf("✓ 任务流打通:Gateway publish (seq=%d) → NATS → Dispatcher consumetask_id=%s", seq, task.ID)
case <-time.After(5 * time.Second):
t.Fatal("timeout: dispatcher 未收到任务")
}
}
// TestToolCallRoundTrip 模拟 Dispatcher 经 NATS 调用 → mcp-go 响应 的工具调用闭环。
func TestToolCallRoundTrip(t *testing.T) {
url := startEmbeddedNATS(t)
// --- mcp-go 侧:以队列组订阅工具主题并响应 ---
srv, err := bus.Connect(url)
if err != nil {
t.Fatalf("mcp connect: %v", err)
}
defer srv.Close()
unsub, err := srv.ServeTool(contract.SubjectToolsGoAll, contract.QueueToolsGo,
func(_ context.Context, call *contract.ToolCall) *contract.ToolResult {
if call.Tool != "wiki_search" {
return &contract.ToolResult{OK: false, Error: "unknown tool"}
}
return &contract.ToolResult{OK: true, Content: "命中:" + call.Args["q"].(string)}
})
if err != nil {
t.Fatalf("serve tool: %v", err)
}
defer func() { _ = unsub() }()
// --- Dispatcher 侧:同步调用工具 ---
dp, err := bus.Connect(url)
if err != nil {
t.Fatalf("dispatcher connect: %v", err)
}
defer dp.Close()
ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
defer cancel()
res, err := dp.CallTool(ctx, contract.ToolSubjectGo("wiki_search"), &contract.ToolCall{
Tool: "wiki_search",
TaskID: "task_tool_001",
Args: map[string]any{"q": "向量检索"},
})
if err != nil {
t.Fatalf("call tool: %v", err)
}
if !res.OK || res.Content != "命中:向量检索" {
t.Fatalf("tool result = %+v, want ok content=命中:向量检索", res)
}
t.Logf("✓ 工具调用闭环:Dispatcher → sundynix.tools.go.wiki_search → mcp-go → %q", res.Content)
}
// TestTokenStreamRoundTrip 模拟 Dispatcher 回流 Token → Gateway 订阅 的流式闭环。
func TestTokenStreamRoundTrip(t *testing.T) {
url := startEmbeddedNATS(t)
// Gateway 侧:先订阅(core NATS 无持久化,须先连)。
gw, err := bus.Connect(url)
if err != nil {
t.Fatalf("gateway connect: %v", err)
}
defer gw.Close()
const taskID = "task_stream_001"
var got []string
done := make(chan struct{})
unsub, err := gw.SubscribeTokens(taskID,
func(tok []byte) { got = append(got, string(tok)) },
func() { close(done) },
)
if err != nil {
t.Fatalf("subscribe tokens: %v", err)
}
defer func() { _ = unsub() }()
// Dispatcher 侧:逐 Token 回流后发结束信号。
dp, err := bus.Connect(url)
if err != nil {
t.Fatalf("dispatcher connect: %v", err)
}
defer dp.Close()
want := []string{"Hello", " ", "Agent", "!"}
for _, tok := range want {
if err := dp.PublishToken(taskID, []byte(tok)); err != nil {
t.Fatalf("publish token: %v", err)
}
}
if err := dp.CompleteStream(taskID); err != nil {
t.Fatalf("complete stream: %v", err)
}
select {
case <-done:
joined := ""
for _, s := range got {
joined += s
}
if joined != "Hello Agent!" {
t.Fatalf("token stream = %q, want %q", joined, "Hello Agent!")
}
t.Logf("✓ Token 流闭环:Dispatcher 回流 %d 个 token → Gateway 拼回 %q", len(got), joined)
case <-time.After(5 * time.Second):
t.Fatal("timeout: 未收到流结束信号")
}
}
// TestConcurrentConsume 验证并发消费:一个长时间阻塞的任务(模拟 HITL 待审)
// 不应阻塞后续任务——旧的串行消费下本测试会超时失败。
func TestConcurrentConsume(t *testing.T) {
url := startEmbeddedNATS(t)
gw, err := bus.Connect(url)
if err != nil {
t.Fatalf("gateway connect: %v", err)
}
defer gw.Close()
ctx, cancel := context.WithTimeout(context.Background(), 15*time.Second)
defer cancel()
if err := gw.EnsureTaskStream(ctx); err != nil {
t.Fatalf("ensure stream: %v", err)
}
dp, err := bus.Connect(url)
if err != nil {
t.Fatalf("dispatcher connect: %v", err)
}
defer dp.Close()
blockA := make(chan struct{})
doneB := make(chan string, 1)
stop, err := dp.ConsumeTasks(ctx, func(_ context.Context, task *contract.Task) error {
switch task.ID {
case "A":
<-blockA // 模拟审批长阻塞
case "B":
doneB <- task.ID
}
return nil
})
if err != nil {
t.Fatalf("consume: %v", err)
}
defer stop()
defer close(blockA)
// 先发 A 并等它被消费、卡在 handler 里;再发 B。
if _, err := gw.PublishTask(ctx, &contract.Task{ID: "A", Graph: json.RawMessage(`{}`)}); err != nil {
t.Fatalf("publish A: %v", err)
}
time.Sleep(400 * time.Millisecond)
if _, err := gw.PublishTask(ctx, &contract.Task{ID: "B", Graph: json.RawMessage(`{}`)}); err != nil {
t.Fatalf("publish B: %v", err)
}
select {
case id := <-doneB:
if id != "B" {
t.Fatalf("unexpected task done: %q", id)
}
t.Log("✓ 并发消费生效:A 仍阻塞时 B 已完成")
case <-time.After(4 * time.Second):
t.Fatal("B 被阻塞的 A 卡住——并发消费未生效(仍是串行)")
}
}