优化自适应扫描系统 & 修复 POC 调度问题

自适应扫描优化:
- target/ceiling 分离,自适应池可向上探索而非锁死在 target
- assessHealth 阈值按网络环境区分(LAN 收紧 / Internet 放宽)
- RTT 漂移时动态压低 target,配合 AIMD 双重降速
- 去掉 semaphore 双层流控,由 ants pool 统一反压
- 探测端口从 3 个扩充到 8 个,减少 RTT 采样偏差
- computeRetries 按环境调整目标概率和上限

Bug 修复:
- AdaptivePool.Wait() 加 10 分钟超时,防止 goroutine 卡死时永久挂起
- CEL 环境初始化失败后允许重试(sync.Once → sync.Mutex + 标志位)
- CAS 自旋加 runtime.Gosched() 退避,减少高并发下 CPU 空转
- -full 模式下 web 插件跳过 IsMarkedWebService 检查 #588
- 不确定服务补做 HTTP 回退探测,覆盖自定义框架漏网场景
- POC sets 纯字面量值跳过 CEL 编译,消除大量误报错误日志
This commit is contained in:
ZacharyZcR
2026-06-14 22:23:48 +08:00
committed by ZacharyZcR
parent c49c23c7f0
commit 8402be98e3
15 changed files with 798 additions and 104 deletions
+549
View File
@@ -0,0 +1,549 @@
package core
import (
"sync/atomic"
"testing"
"time"
)
// =============================================================================
// 优化 1target/ceiling 分离
// =============================================================================
func TestOpt1_TargetCeilingSeparation_TuneConfig(t *testing.T) {
config := makeDefaultConfig()
session := makeTestSession(config)
ep := &EnvironmentProfile{
Net: NetworkProfile{
Env: EnvLAN,
RTTMedian: 1 * time.Millisecond,
RTTStddev: 500 * time.Microsecond,
LossRate: 0.0,
Samples: 30,
},
System: SystemProfile{FDLimit: 65536, NumCPU: 8},
}
ep.TuneConfig(config, session)
if config.ThreadCeiling <= 0 {
t.Fatalf("ThreadCeiling 未被设置: %d", config.ThreadCeiling)
}
// 内网 factor=1.5,非显式 → target=ceiling=recommended
// 但 ceiling 应该 >= target
if config.ThreadCeiling < config.ThreadNum {
t.Errorf("Ceiling(%d) < ThreadNum(%d)", config.ThreadCeiling, config.ThreadNum)
}
t.Logf("target=%d, ceiling=%d", config.ThreadNum, config.ThreadCeiling)
}
func TestOpt1_TargetCeilingSeparation_ExplicitT(t *testing.T) {
config := makeDefaultConfig()
config.ThreadNum = 200
config.ThreadNumExplicit = true
session := makeTestSession(config)
ep := &EnvironmentProfile{
Net: NetworkProfile{
Env: EnvInternet,
RTTMedian: 100 * time.Millisecond,
RTTStddev: 30 * time.Millisecond,
LossRate: 0.0,
Samples: 20,
},
System: SystemProfile{FDLimit: 65536, NumCPU: 8},
}
ep.TuneConfig(config, session)
// 用户显式指定 -t → ceiling = threadNum = 200
if config.ThreadCeiling != 200 {
t.Errorf("显式 -t 200: ceiling=%d, want 200", config.ThreadCeiling)
}
if config.ThreadNum != 200 {
t.Errorf("显式 -t 200: threadNum=%d, want 200", config.ThreadNum)
}
}
func TestOpt1_PoolUsesCeiling(t *testing.T) {
metrics := &ScanMetrics{}
target, ceiling := 50, 200
pool, err := NewAdaptivePool(target, ceiling, func(interface{}) {}, metrics)
if err != nil {
t.Fatalf("创建池失败: %v", err)
}
defer pool.Release()
pool.inSlowStart = false
pool.tune(target)
// 注入健康 metrics 让池增长
for i := 0; i < 200; i++ {
metrics.RecordConnect(time.Millisecond)
}
// 多次 adjust,池应能增长超过 target 但不超过 ceiling
for i := 0; i < 30; i++ {
pool.lastCheck.Store(0)
pool.adjust()
}
finalCap := pool.Cap()
if finalCap <= target {
t.Errorf("池应能超过 target(%d): cap=%d", target, finalCap)
}
if finalCap > ceiling {
t.Errorf("池不应超过 ceiling(%d): cap=%d", ceiling, finalCap)
}
t.Logf("target=%d, ceiling=%d, finalCap=%d", target, ceiling, finalCap)
}
func TestOpt1_FDLimitConstraintsBothFields(t *testing.T) {
config := makeDefaultConfig()
config.ThreadNum = 1000
session := makeTestSession(config)
ep := &EnvironmentProfile{
Net: NetworkProfile{
Env: EnvLAN,
RTTMedian: 1 * time.Millisecond,
RTTStddev: 500 * time.Microsecond,
LossRate: 0.0,
Samples: 30,
},
System: SystemProfile{FDLimit: 256, NumCPU: 4},
}
ep.TuneConfig(config, session)
maxFD := 256 * 6 / 10
if config.ThreadNum > maxFD {
t.Errorf("ThreadNum(%d) 超过 fd 限制(%d)", config.ThreadNum, maxFD)
}
if config.ThreadCeiling > maxFD {
t.Errorf("ThreadCeiling(%d) 超过 fd 限制(%d)", config.ThreadCeiling, maxFD)
}
}
// =============================================================================
// 优化 2RTT 漂移微调 target
// =============================================================================
func TestOpt2_RTTDriftReducesTarget(t *testing.T) {
metrics := &ScanMetrics{}
pool, err := NewAdaptivePool(200, 400, func(interface{}) {}, metrics)
if err != nil {
t.Fatalf("创建池失败: %v", err)
}
defer pool.Release()
pool.inSlowStart = false
pool.tune(200)
// 建立基线:slow EMA 锚定在 1ms 附近
for i := 0; i < 500; i++ {
metrics.RecordConnect(1 * time.Millisecond)
}
origTarget := atomic.LoadInt32(&pool.target)
// RTT 突增到 100ms100 倍),大量喂入让 fast EMA 拉开差距
for i := 0; i < 1000; i++ {
metrics.RecordConnect(100 * time.Millisecond)
}
ratio := metrics.RTTRatio()
t.Logf("RTT ratio after spike: %.2f", ratio)
if ratio <= 3.0 {
t.Skipf("RTT ratio=%.2fEMA 差距不够大,跳过", ratio)
}
// 需要足够的新 metrics 让 assessHealth 的 deltaTotal >= 30
for i := 0; i < 50; i++ {
metrics.RecordConnect(100 * time.Millisecond)
}
// 多次 adjust 触发 maybeReduceTarget
for i := 0; i < 10; i++ {
pool.lastCheck.Store(0)
pool.prevSnapshot = MetricsSnapshot{} // 重置快照让 delta 足够
pool.adjust()
}
newTarget := atomic.LoadInt32(&pool.target)
if newTarget >= origTarget {
t.Errorf("RTT 漂移后 target 应降低: %d -> %d (ratio=%.2f)", origTarget, newTarget, ratio)
}
// 不应低于 ceiling/5
minTarget := atomic.LoadInt32(&pool.ceiling) / 5
if newTarget < minTarget {
t.Errorf("target(%d) 低于下限(%d)", newTarget, minTarget)
}
t.Logf("RTT drift: ratio=%.2f, target %d -> %d (min=%d)", ratio, origTarget, newTarget, minTarget)
}
func TestOpt2_NoReductionWhenStable(t *testing.T) {
metrics := &ScanMetrics{}
pool, err := NewAdaptivePool(200, 400, func(interface{}) {}, metrics)
if err != nil {
t.Fatalf("创建池失败: %v", err)
}
defer pool.Release()
pool.inSlowStart = false
pool.tune(200)
// 稳定 RTT
for i := 0; i < 200; i++ {
metrics.RecordConnect(10 * time.Millisecond)
}
origTarget := atomic.LoadInt32(&pool.target)
for i := 0; i < 10; i++ {
pool.lastCheck.Store(0)
pool.adjust()
}
newTarget := atomic.LoadInt32(&pool.target)
if newTarget != origTarget {
t.Errorf("稳定 RTT 不应改变 target: %d -> %d", origTarget, newTarget)
}
}
// =============================================================================
// 优化 3assessHealth 阈值跟 NetworkEnv 关联
// =============================================================================
func TestOpt3_LANTighterThresholds(t *testing.T) {
metrics := &ScanMetrics{}
pool, err := NewAdaptivePool(100, 100, func(interface{}) {}, metrics, EnvLAN)
if err != nil {
t.Fatalf("创建池失败: %v", err)
}
defer pool.Release()
pool.inSlowStart = false
pool.tune(100)
// 10% exhaust rate — 对 LAN 来说应该是 Congested(阈值 8%
for i := 0; i < 100; i++ {
if i < 10 {
metrics.RecordExhausted()
} else {
metrics.RecordConnect(time.Millisecond)
}
}
pool.lastCheck.Store(0)
pool.adjust()
if pool.Cap() >= 100 {
t.Errorf("LAN 10%% exhaust 应触发降速: cap=%d", pool.Cap())
}
t.Logf("LAN tight threshold: cap=%d (from 100)", pool.Cap())
}
func TestOpt3_InternetLooseThresholds(t *testing.T) {
metrics := &ScanMetrics{}
pool, err := NewAdaptivePool(100, 100, func(interface{}) {}, metrics, EnvInternet)
if err != nil {
t.Fatalf("创建池失败: %v", err)
}
defer pool.Release()
pool.inSlowStart = false
pool.tune(100)
// 10% exhaust rate — 对 Internet 来说不算 Congested(阈值 25%),应是 Stressed
for i := 0; i < 100; i++ {
if i < 10 {
metrics.RecordExhausted()
} else {
metrics.RecordConnect(10 * time.Millisecond)
}
}
pool.lastCheck.Store(0)
pool.adjust()
capAfter := pool.Cap()
// Internet 对 10% exhaust 只是 Stressed(×0.85),不是 Congested(×0.5
if capAfter < 80 {
t.Errorf("Internet 10%% exhaust 不应大幅降速: cap=%d", capAfter)
}
t.Logf("Internet loose threshold: cap=%d (from 100)", capAfter)
}
func TestOpt3_EnvAffectsHealthDecision(t *testing.T) {
envs := []struct {
env NetworkEnv
name string
}{
{EnvLAN, "LAN"},
{EnvWAN, "WAN"},
{EnvInternet, "Internet"},
}
var caps []int
for _, e := range envs {
metrics := &ScanMetrics{}
pool, err := NewAdaptivePool(100, 100, func(interface{}) {}, metrics, e.env)
if err != nil {
t.Fatalf("创建池失败: %v", err)
}
pool.inSlowStart = false
pool.tune(100)
// 相同的 12% exhaust rate
for i := 0; i < 100; i++ {
if i < 12 {
metrics.RecordExhausted()
} else {
metrics.RecordConnect(time.Millisecond)
}
}
pool.lastCheck.Store(0)
pool.adjust()
caps = append(caps, pool.Cap())
pool.Release()
t.Logf("%s: cap=%d (12%% exhaust)", e.name, caps[len(caps)-1])
}
// LAN 反应最激烈(cap 最低),Internet 最宽容(cap 最高)
if caps[0] >= caps[2] {
t.Errorf("LAN cap(%d) 应 < Internet cap(%d) for same exhaust rate", caps[0], caps[2])
}
}
// =============================================================================
// 优化 4:去掉 semaphoreants 池天然反压
// =============================================================================
func TestOpt4_SemaphoreRemoved(t *testing.T) {
// 验证 portScanTask 结构体不再有 semaphore 字段
// 如果 semaphore 被加回来,这段代码编译就会报 "unknown field"
_ = portScanTask{
host: "127.0.0.1",
port: 80,
addr: "127.0.0.1:80",
}
t.Log("portScanTask 无 semaphore 字段,反压由 ants pool 统一管理")
}
// =============================================================================
// 优化 5:扩充探测端口
// =============================================================================
func TestOpt5_ProbePortsExpanded(t *testing.T) {
if len(probePorts) < 5 {
t.Errorf("probePorts 只有 %d 个,应该扩充到至少 5 个", len(probePorts))
}
// 验证包含关键端口
required := map[int]bool{80: false, 443: false, 22: false}
for _, p := range probePorts {
if _, ok := required[p]; ok {
required[p] = true
}
}
for port, found := range required {
if !found {
t.Errorf("probePorts 缺少关键端口 %d", port)
}
}
// 验证没有重复
seen := make(map[int]bool)
for _, p := range probePorts {
if seen[p] {
t.Errorf("probePorts 有重复端口 %d", p)
}
seen[p] = true
}
t.Logf("probePorts = %v (%d 个)", probePorts, len(probePorts))
}
// =============================================================================
// 优化 6computeRetries 环境自适应
// =============================================================================
func TestOpt6_RetriesEnvAware(t *testing.T) {
lossRate := 0.3 // 30% 丢包
lanRetry := computeRetries(lossRate, EnvLAN)
wanRetry := computeRetries(lossRate, EnvWAN)
inetRetry := computeRetries(lossRate, EnvInternet)
// LAN 目标概率更严格(0.5%),应该重试更多;但上限更低(4)
// Internet 目标概率更宽松(2%),应该重试更少;但上限更高(6)
t.Logf("30%% loss: LAN=%d, WAN=%d, Internet=%d", lanRetry, wanRetry, inetRetry)
if lanRetry < 1 || lanRetry > 4 {
t.Errorf("LAN retry=%d, 应在 [1,4]", lanRetry)
}
if wanRetry < 1 || wanRetry > 5 {
t.Errorf("WAN retry=%d, 应在 [1,5]", wanRetry)
}
if inetRetry < 1 || inetRetry > 6 {
t.Errorf("Internet retry=%d, 应在 [1,6]", inetRetry)
}
}
func TestOpt6_RetriesMaxByEnv(t *testing.T) {
// 高丢包率,各环境应返回各自上限
lanMax := computeRetries(0.99, EnvLAN)
wanMax := computeRetries(0.99, EnvWAN)
inetMax := computeRetries(0.99, EnvInternet)
if lanMax != 4 {
t.Errorf("LAN max retry=%d, want 4", lanMax)
}
if wanMax != 5 {
t.Errorf("WAN max retry=%d, want 5", wanMax)
}
if inetMax != 6 {
t.Errorf("Internet max retry=%d, want 6", inetMax)
}
}
func TestOpt6_RetriesMathCorrectness(t *testing.T) {
envs := []struct {
env NetworkEnv
targetProb float64
name string
}{
{EnvLAN, 0.005, "LAN"},
{EnvWAN, 0.01, "WAN"},
{EnvInternet, 0.02, "Internet"},
}
for _, e := range envs {
for _, loss := range []float64{0.05, 0.10, 0.20, 0.30} {
retries := computeRetries(loss, e.env)
prob := 1.0
for i := 0; i < retries; i++ {
prob *= loss
}
// 重试后全失败概率应 < targetProb(除非被 clamp 了)
if prob >= e.targetProb && retries < 4 {
t.Errorf("%s loss=%.0f%% retries=%d: P=%.6f >= %.3f",
e.name, loss*100, retries, prob, e.targetProb)
}
}
}
}
// =============================================================================
// 端到端集成:全链路验证
// =============================================================================
func TestOptAll_EndToEnd_LANToPool(t *testing.T) {
// 模拟内网探测 → TuneConfig → 创建池 → 池根据 env 自适应
profile := classifyNetwork(
makeDurations([]int{1, 1, 2, 2, 2, 3, 3, 3, 4, 5}),
0, 10,
)
config := makeDefaultConfig()
session := makeTestSession(config)
ep := &EnvironmentProfile{Net: *profile, System: SystemProfile{FDLimit: 65536, NumCPU: 8}}
ep.TuneConfig(config, session)
// 验证 env 被存储
if config.DetectedNetworkEnv != int(EnvLAN) {
t.Errorf("DetectedNetworkEnv=%d, want %d(LAN)", config.DetectedNetworkEnv, int(EnvLAN))
}
// 验证 ceiling 合理
if config.ThreadCeiling < config.ThreadNum {
t.Errorf("ceiling(%d) < target(%d)", config.ThreadCeiling, config.ThreadNum)
}
// 创建池并验证 env 传递
netEnv := NetworkEnv(config.DetectedNetworkEnv)
metrics := &ScanMetrics{}
pool, err := NewAdaptivePool(config.ThreadNum, config.ThreadCeiling, func(interface{}) {}, metrics, netEnv)
if err != nil {
t.Fatalf("创建池失败: %v", err)
}
defer pool.Release()
if pool.networkEnv != EnvLAN {
t.Errorf("池的 networkEnv=%v, want LAN", pool.networkEnv)
}
t.Logf("端到端 LAN: target=%d ceiling=%d env=%v maxRetry=%d",
config.ThreadNum, config.ThreadCeiling, netEnv, config.MaxRetries)
}
func TestOptAll_EndToEnd_InternetToPool(t *testing.T) {
profile := classifyNetwork(
makeDurations([]int{60, 70, 80, 90, 100, 110, 120, 130, 140, 150}),
0, 10,
)
config := makeDefaultConfig()
session := makeTestSession(config)
ep := &EnvironmentProfile{Net: *profile, System: SystemProfile{FDLimit: 4096, NumCPU: 4}}
ep.TuneConfig(config, session)
if config.DetectedNetworkEnv != int(EnvInternet) {
t.Errorf("DetectedNetworkEnv=%d, want %d(Internet)", config.DetectedNetworkEnv, int(EnvInternet))
}
// 公网 target 应明显低于默认 600
if config.ThreadNum >= 600 {
t.Errorf("公网 threadNum=%d, 应 < 600", config.ThreadNum)
}
// ceiling 应 == target(非显式模式)
if config.ThreadCeiling != config.ThreadNum {
t.Errorf("非显式模式 ceiling(%d) != target(%d)", config.ThreadCeiling, config.ThreadNum)
}
netEnv := NetworkEnv(config.DetectedNetworkEnv)
metrics := &ScanMetrics{}
pool, err := NewAdaptivePool(config.ThreadNum, config.ThreadCeiling, func(interface{}) {}, metrics, netEnv)
if err != nil {
t.Fatalf("创建池失败: %v", err)
}
defer pool.Release()
// 注入 12% exhaustInternet 环境应只是 Stressed 而不是 Congested
pool.inSlowStart = false
pool.tune(config.ThreadNum)
for i := 0; i < 100; i++ {
if i < 12 {
metrics.RecordExhausted()
} else {
metrics.RecordConnect(80 * time.Millisecond)
}
}
pool.lastCheck.Store(0)
pool.adjust()
// cap 不应被砍到一半以下(Stressed 只降 15%
if pool.Cap() < config.ThreadNum*7/10 {
t.Errorf("Internet 12%% exhaust 降速过猛: %d -> %d", config.ThreadNum, pool.Cap())
}
t.Logf("端到端 Internet: target=%d ceiling=%d cap_after_stress=%d",
config.ThreadNum, config.ThreadCeiling, pool.Cap())
}