354 lines
11 KiB
JavaScript
354 lines
11 KiB
JavaScript
'use strict'
|
||
|
||
const fs = require('fs')
|
||
const http = require('http')
|
||
const path = require('path')
|
||
const { EventEmitter } = require('events')
|
||
const { spawn } = require('child_process')
|
||
|
||
const STATE_LABEL = {
|
||
stopped: '已停止',
|
||
starting: '启动中',
|
||
running: '运行中',
|
||
stopping: '停止中',
|
||
crashed: '异常退出'
|
||
}
|
||
|
||
// 优雅关闭握手：POST server.shutdownPath，让服务端自己走 signal.NotifyContext
|
||
// 的关闭流程；请求失败或接口未实现时回退到进程信号。
|
||
const DEFAULT_SHUTDOWN_PATH = '/api/shutdown'
|
||
const SHUTDOWN_REQUEST_TIMEOUT_MS = 2000
|
||
// 服务端 srv.Shutdown 的超时是 10s，客户端必须留足余量再强杀
|
||
const DEFAULT_SHUTDOWN_TIMEOUT_MS = 12000
|
||
// 握手不可用时信号已能结束进程，不必再等满超时
|
||
const FALLBACK_FORCE_KILL_MS = 1500
|
||
// 连续健康这么久才认为崩溃计数可以清零
|
||
const HEALTHY_RESET_MS = 60000
|
||
|
||
// 服务端守护：拉起 scoreboard.exe、转发日志、健康探测、崩溃自动重启。
|
||
// 现场无人值守，所以「崩了自动拉起」比「暴露复杂控制」更重要。
|
||
class ServiceSupervisor extends EventEmitter {
|
||
constructor({ values, exePath, log }) {
|
||
super()
|
||
this.values = values
|
||
this.server = values.server
|
||
this.exePath = exePath
|
||
this.log = log
|
||
|
||
this.child = null
|
||
this.state = 'stopped'
|
||
this.healthy = false
|
||
this.pid = null
|
||
this.restarts = 0
|
||
this.startedAt = null
|
||
this.lastError = null
|
||
|
||
this.stopRequested = false
|
||
this.healthySince = null
|
||
this.healthTimer = null
|
||
this.restartTimer = null
|
||
this.forceKillTimer = null
|
||
}
|
||
|
||
get status() {
|
||
const host = this.healthHost()
|
||
return {
|
||
state: this.state,
|
||
stateLabel: STATE_LABEL[this.state] || this.state,
|
||
healthy: this.healthy,
|
||
pid: this.pid,
|
||
restarts: this.restarts,
|
||
startedAt: this.startedAt,
|
||
lastError: this.lastError,
|
||
exePath: this.exePath,
|
||
url: `http://${host}:${this.server.port}`,
|
||
adminUrl: `http://${host}:${this.server.port}/admin`
|
||
}
|
||
}
|
||
|
||
// 服务端可能监听 0.0.0.0，本机探测要走回环地址
|
||
healthHost() {
|
||
const host = this.server.host || '127.0.0.1'
|
||
return host === '0.0.0.0' || host === '::' ? '127.0.0.1' : host
|
||
}
|
||
|
||
emitStatus() {
|
||
this.emit('status', this.status)
|
||
}
|
||
|
||
setState(state, patch = {}) {
|
||
const changed = this.state !== state
|
||
this.state = state
|
||
Object.assign(this, patch)
|
||
if (changed) {
|
||
this.log.push('console', `服务状态：${STATE_LABEL[state] || state}${this.lastError ? `（${this.lastError}）` : ''}`)
|
||
}
|
||
this.emitStatus()
|
||
}
|
||
|
||
start() {
|
||
if (this.child || this.state === 'starting' || this.state === 'running') return this.status
|
||
this.stopRequested = false
|
||
|
||
if (!this.exePath || !fs.existsSync(this.exePath)) {
|
||
this.setState('crashed', {
|
||
lastError: `未找到服务端程序：${this.exePath || this.server.exe}`
|
||
})
|
||
return this.status
|
||
}
|
||
|
||
const args = Array.isArray(this.server.args) ? this.server.args : []
|
||
// 工作目录取 exe 所在目录：部署包里 config.yaml / data/ 都跟 exe 平级
|
||
const cwd = path.dirname(this.exePath)
|
||
|
||
this.lastError = null
|
||
this.healthySince = null
|
||
this.startedAt = new Date().toISOString()
|
||
this.setState('starting')
|
||
this.log.push('console', `启动服务端：${this.exePath} ${args.join(' ')}（cwd=${cwd}）`)
|
||
|
||
let child
|
||
try {
|
||
child = spawn(this.exePath, args, {
|
||
cwd,
|
||
windowsHide: true,
|
||
stdio: ['ignore', 'pipe', 'pipe']
|
||
})
|
||
} catch (err) {
|
||
this.setState('crashed', { lastError: `启动失败：${err.message}` })
|
||
this.scheduleRestart()
|
||
return this.status
|
||
}
|
||
|
||
this.child = child
|
||
this.pid = child.pid
|
||
|
||
if (child.stdout) {
|
||
child.stdout.setEncoding('utf8')
|
||
child.stdout.on('data', (chunk) => this.log.append('server', chunk))
|
||
}
|
||
if (child.stderr) {
|
||
child.stderr.setEncoding('utf8')
|
||
child.stderr.on('data', (chunk) => this.log.append('server!', chunk))
|
||
}
|
||
|
||
child.on('error', (err) => {
|
||
this.log.push('console', `服务端进程错误：${err.message}`)
|
||
this.setState('crashed', { lastError: err.message })
|
||
})
|
||
|
||
child.on('exit', (code, signal) => this.handleExit(code, signal))
|
||
|
||
this.setState('running')
|
||
this.startHealthCheck()
|
||
return this.status
|
||
}
|
||
|
||
handleExit(code, signal) {
|
||
if (this.forceKillTimer) {
|
||
clearTimeout(this.forceKillTimer)
|
||
this.forceKillTimer = null
|
||
}
|
||
this.stopHealthCheck()
|
||
this.child = null
|
||
this.pid = null
|
||
this.healthy = false
|
||
this.healthySince = null
|
||
|
||
const reason = signal ? `信号 ${signal}` : `退出码 ${code}`
|
||
if (this.stopRequested) {
|
||
this.log.push('console', `服务端已停止（${reason}）`)
|
||
this.setState('stopped', { lastError: null })
|
||
return
|
||
}
|
||
|
||
this.setState('crashed', { lastError: `服务端意外退出（${reason}）` })
|
||
this.scheduleRestart()
|
||
}
|
||
|
||
scheduleRestart() {
|
||
const max = Number(this.server.maxRestarts) || 0
|
||
if (max > 0 && this.restarts >= max) {
|
||
this.log.push('console', `已达到最大重启次数 ${max}，不再自动拉起`)
|
||
return
|
||
}
|
||
const delay = (Number(this.server.restartDelayMs) || 3000) * Math.min(this.restarts + 1, 5)
|
||
this.restarts += 1
|
||
this.log.push('console', `${Math.round(delay / 1000)} 秒后自动重启服务端（第 ${this.restarts} 次）`)
|
||
this.emitStatus()
|
||
if (this.restartTimer) clearTimeout(this.restartTimer)
|
||
this.restartTimer = setTimeout(() => {
|
||
this.restartTimer = null
|
||
if (!this.stopRequested) this.start()
|
||
}, delay)
|
||
}
|
||
|
||
stop() {
|
||
return new Promise((resolve) => {
|
||
if (this.restartTimer) {
|
||
clearTimeout(this.restartTimer)
|
||
this.restartTimer = null
|
||
}
|
||
const child = this.child
|
||
if (!child) {
|
||
this.setState('stopped', { lastError: null })
|
||
resolve()
|
||
return
|
||
}
|
||
|
||
this.stopRequested = true
|
||
this.setState('stopping')
|
||
this.log.push('console', '正在停止服务端…')
|
||
|
||
child.once('exit', () => resolve())
|
||
|
||
const timeoutMs = Number(this.server.shutdownTimeoutMs) || DEFAULT_SHUTDOWN_TIMEOUT_MS
|
||
// 先按最坏情况挂强杀兜底，握手成功后也不撤：服务端可能卡在关闭流程里
|
||
this.armForceKill(timeoutMs)
|
||
|
||
this.requestShutdown().then((accepted) => {
|
||
if (this.child !== child) return
|
||
if (accepted) {
|
||
this.log.push('console', `服务端已接受关闭请求，最多等待 ${Math.round(timeoutMs / 1000)} 秒`)
|
||
return
|
||
}
|
||
// 握手不可用：服务端还没实现该接口，或进程已无响应。
|
||
// 退回进程信号；Windows 上 Node 的 kill() 等同强杀，所以这条路只是兜底，
|
||
// 真正的优雅关闭依赖服务端补上 shutdownPath 接口。
|
||
this.log.push('console', '关闭接口不可用，退回进程信号')
|
||
this.armForceKill(FALLBACK_FORCE_KILL_MS)
|
||
try {
|
||
child.kill()
|
||
} catch (err) {
|
||
this.log.push('console', `发送停止信号失败：${err.message}`)
|
||
}
|
||
})
|
||
})
|
||
}
|
||
|
||
// 优雅关闭握手：让服务端自己走 signal.NotifyContext → srv.Shutdown 的流程，
|
||
// 10s 内落盘退出。接口按回环地址访问、不鉴权，只在本机有意义。
|
||
// 依赖：服务端实现 POST <server.shutdownPath>（2xx 表示已受理）。
|
||
requestShutdown() {
|
||
return new Promise((resolve) => {
|
||
const body = Buffer.from('{}')
|
||
const request = http.request(
|
||
{
|
||
host: this.healthHost(),
|
||
port: this.server.port,
|
||
path: this.server.shutdownPath || DEFAULT_SHUTDOWN_PATH,
|
||
method: 'POST',
|
||
timeout: SHUTDOWN_REQUEST_TIMEOUT_MS,
|
||
headers: {
|
||
'Content-Type': 'application/json',
|
||
'Content-Length': body.length,
|
||
Connection: 'close'
|
||
}
|
||
},
|
||
(response) => {
|
||
response.resume()
|
||
const accepted = response.statusCode >= 200 && response.statusCode < 300
|
||
if (!accepted) this.log.push('console', `关闭接口返回 HTTP ${response.statusCode}`)
|
||
resolve(accepted)
|
||
}
|
||
)
|
||
request.on('timeout', () => request.destroy(new Error('请求超时')))
|
||
request.on('error', (err) => {
|
||
this.log.push('console', `关闭接口请求失败：${err.message}`)
|
||
resolve(false)
|
||
})
|
||
request.end(body)
|
||
})
|
||
}
|
||
|
||
armForceKill(ms) {
|
||
if (this.forceKillTimer) clearTimeout(this.forceKillTimer)
|
||
this.forceKillTimer = setTimeout(() => {
|
||
this.forceKillTimer = null
|
||
if (!this.child || !this.pid) return
|
||
this.log.push('console', `服务端未在 ${Math.round(ms / 1000)} 秒内退出，强制结束`)
|
||
try {
|
||
this.child.kill('SIGKILL')
|
||
} catch {
|
||
// 进程可能已经没了
|
||
}
|
||
}, ms)
|
||
}
|
||
|
||
async restart() {
|
||
await this.stop()
|
||
this.restarts = 0
|
||
return this.start()
|
||
}
|
||
|
||
startHealthCheck() {
|
||
this.stopHealthCheck()
|
||
const interval = Number(this.server.healthIntervalMs) || 5000
|
||
const tick = () => this.probe()
|
||
setTimeout(tick, 800)
|
||
this.healthTimer = setInterval(tick, interval)
|
||
}
|
||
|
||
stopHealthCheck() {
|
||
if (this.healthTimer) {
|
||
clearInterval(this.healthTimer)
|
||
this.healthTimer = null
|
||
}
|
||
}
|
||
|
||
probe() {
|
||
if (!this.child) return
|
||
const request = http.get(
|
||
{
|
||
host: this.healthHost(),
|
||
port: this.server.port,
|
||
path: '/api/health',
|
||
timeout: 2500
|
||
},
|
||
(response) => {
|
||
response.resume()
|
||
const ok = response.statusCode === 200
|
||
this.setHealthy(ok)
|
||
}
|
||
)
|
||
request.on('timeout', () => request.destroy())
|
||
request.on('error', () => this.setHealthy(false))
|
||
}
|
||
|
||
setHealthy(ok) {
|
||
const changed = this.healthy !== ok
|
||
if (ok) {
|
||
if (!this.healthySince) this.healthySince = Date.now()
|
||
} else {
|
||
this.healthySince = null
|
||
}
|
||
this.healthy = ok
|
||
if (changed) {
|
||
this.log.push('console', ok ? '服务端健康检查通过' : '服务端健康检查失败（等待重试）')
|
||
}
|
||
|
||
// 连续健康一段时间后清零崩溃计数：否则长时间赛事里历史崩溃会耗尽
|
||
// maxRestarts 配额，之后真崩了反而不自动拉起
|
||
let reset = false
|
||
if (ok && this.restarts > 0) {
|
||
const resetAfter = Number(this.server.healthyResetMs) || HEALTHY_RESET_MS
|
||
if (Date.now() - this.healthySince >= resetAfter) {
|
||
this.restarts = 0
|
||
reset = true
|
||
this.log.push('console', '服务端持续健康，崩溃计数已清零')
|
||
}
|
||
}
|
||
|
||
if (changed || reset) this.emitStatus()
|
||
}
|
||
|
||
dispose() {
|
||
if (this.restartTimer) clearTimeout(this.restartTimer)
|
||
if (this.forceKillTimer) clearTimeout(this.forceKillTimer)
|
||
this.stopHealthCheck()
|
||
this.removeAllListeners()
|
||
}
|
||
}
|
||
|
||
module.exports = { ServiceSupervisor, STATE_LABEL }
|