Files
HomeAgent/internal/network/monitor.go
root bc9ef15eb0 三层回退恢复机制(L0写前留档/L1恢复梯子/L2离线回滚)+ guard 父守护
- L0: files 插件写受保护系统路径(/etc 等)前自动留档,AbstractBeforeWrite 到 data/file_baseline
- L1: failback 受限 worker 执行恢复梯子 probe→还原DNS/proxy→还原LLM配置+ReloadFromConfig→probe,N轮有界
- L2: tracker changeset 持久化原文 blob,guard 离线 RollbackFromDisk 回滚 agentfs;SystemSnapshot 支撑
- guard 父守护: 心跳 IPC(PING/ACK unix socket, 文件心跳回退)、失败计数、退出码协议(42/43/44)、最后手段
- 发行版路径适配: system.protected_paths/network_paths 可注入,默认面向主流 Linux
- Windows 兼容: guard.go/failback.go 加 //go:build linux, guard_windows.go 提供 no-op 桩
- 修复: guard.yaml last_resort 键冲突、changeset Content 不落盘导致离线回滚丢原文

Build 全绿, vet 干净, system/recovery/ipc/tracker 单元测试全过
2026-08-05 16:00:08 +08:00

222 lines
4.5 KiB
Go

package network
import (
"context"
"fmt"
"net"
"net/http"
"sync"
"time"
"gitcode.com/JianFeeeee/HomeAgent/pkg/types"
)
type Monitor struct {
mu sync.RWMutex
client *http.Client
interval time.Duration
endpoints []string
status []EndpointStatus
}
type EndpointStatus struct {
URL string
Reachable bool
Latency time.Duration
LastCheck time.Time
Error string
}
func NewMonitor(interval time.Duration) *Monitor {
return &Monitor{
client: &http.Client{
Timeout: 10 * time.Second,
Transport: &http.Transport{
DialContext: (&net.Dialer{
Timeout: 5 * time.Second,
KeepAlive: 30 * time.Second,
}).DialContext,
TLSHandshakeTimeout: 5 * time.Second,
ResponseHeaderTimeout: 5 * time.Second,
DisableKeepAlives: false,
MaxIdleConns: 2,
IdleConnTimeout: 30 * time.Second,
},
},
interval: interval,
status: make([]EndpointStatus, 0),
}
}
func (m *Monitor) Start(ctx context.Context, endpoints []string) {
m.mu.Lock()
m.endpoints = endpoints
m.status = make([]EndpointStatus, len(endpoints))
for i, ep := range endpoints {
m.status[i] = EndpointStatus{URL: ep, Reachable: false}
}
m.mu.Unlock()
ticker := time.NewTicker(m.interval)
defer ticker.Stop()
m.checkAll(ctx)
for {
select {
case <-ticker.C:
m.checkAll(ctx)
case <-ctx.Done():
return
}
}
}
func (m *Monitor) CheckOnce(ctx context.Context, endpoint string) EndpointStatus {
start := time.Now()
req, err := http.NewRequestWithContext(ctx, "HEAD", endpoint, nil)
if err != nil {
return EndpointStatus{URL: endpoint, Reachable: false, Error: err.Error(), LastCheck: time.Now()}
}
resp, err := m.client.Do(req)
latency := time.Since(start)
if err != nil {
return EndpointStatus{URL: endpoint, Reachable: false, Latency: latency, Error: err.Error(), LastCheck: time.Now()}
}
resp.Body.Close()
return EndpointStatus{
URL: endpoint,
Reachable: resp.StatusCode < 500,
Latency: latency,
LastCheck: time.Now(),
}
}
func (m *Monitor) checkAll(ctx context.Context) {
m.mu.RLock()
endpoints := m.endpoints
m.mu.RUnlock()
var wg sync.WaitGroup
results := make([]EndpointStatus, len(endpoints))
for i, ep := range endpoints {
wg.Add(1)
go func(idx int, url string) {
defer wg.Done()
results[idx] = m.CheckOnce(ctx, url)
}(i, ep)
}
wg.Wait()
m.mu.Lock()
m.status = results
m.mu.Unlock()
}
func (m *Monitor) Status() []EndpointStatus {
m.mu.RLock()
defer m.mu.RUnlock()
status := make([]EndpointStatus, len(m.status))
copy(status, m.status)
return status
}
func (m *Monitor) AllReachable() bool {
m.mu.RLock()
defer m.mu.RUnlock()
for _, s := range m.status {
if !s.Reachable {
return false
}
}
return len(m.status) > 0
}
func (m *Monitor) AggregateResult() types.NetworkCheckResult {
m.mu.RLock()
defer m.mu.RUnlock()
result := types.NetworkCheckResult{
LLMAPIReachable: true,
EndpointsConfigured: len(m.status) > 0,
DNSResolving: true,
TCPReachable: true,
}
var totalLatency time.Duration
checked := 0
if !result.EndpointsConfigured {
// 无任何探活端点:不谎报"可达",标为未配置
result.LLMAPIReachable = false
result.Error = "no LLM endpoints configured for health check"
}
for _, s := range m.status {
if !s.Reachable {
result.LLMAPIReachable = false
result.Error = fmt.Sprintf("endpoint %s unreachable: %s", s.URL, s.Error)
}
if s.Latency > 0 {
totalLatency += s.Latency
checked++
}
}
if checked > 0 {
result.Latency = totalLatency / time.Duration(checked)
}
// 延迟阈值检测:平均延迟 > 5s 标记为降级
if result.Latency > 5*time.Second {
result.LatencyDegraded = true
if result.Error == "" {
result.Error = fmt.Sprintf("high latency: %v", result.Latency)
}
}
// DNS 多目标检测
result.DNSResolving = m.checkDNSMulti()
// TCP 拨测:检测基础网络通畅性
result.TCPReachable = m.checkTCPReachability()
return result
}
func (m *Monitor) checkDNSMulti() bool {
targets := []string{"google.com", "baidu.com", "cloudflare.com"}
for _, target := range targets {
_, err := net.LookupHost(target)
if err == nil {
return true
}
}
return false
}
func (m *Monitor) checkTCPReachability() bool {
targets := []struct {
host string
port string
}{
{"8.8.8.8", "53"},
{"1.1.1.1", "53"},
{"208.67.222.222", "53"},
}
for _, t := range targets {
conn, err := net.DialTimeout("tcp", net.JoinHostPort(t.host, t.port), 3*time.Second)
if err == nil {
conn.Close()
return true
}
}
return false
}