命令行与文本处理 / 系统课时
实现远程并行巡检
实现远程并行巡检
本课把单机采集扩展为可控的批量远程巡检。你将使用 PowerShell Remoting 创建并行任务,通过节流限制并发,以超时回收卡住的任务,并让每条结果始终携带批次编号和目标主机名。
学习目标
- 使用
Invoke-Command -AsJob并行采集多台主机。 - 用节流、连接超时和批次等待上限控制资源。
- 为成功、失败和超时主机生成字段一致的结果。
设计并行巡检链路
批量巡检不能简单地同时启动所有命令。合理链路应包含目标队列、并发窗口、远程执行、超时回收和结果归一化。ThrottleLimit 控制同时运行的远程操作数量,批次编号与主机名负责把乱序完成的结果重新关联。

远程执行前应确认名称解析、WinRM、身份验证和防火墙规则已经配置。脚本不要硬编码管理员密码,也不要把 TrustedHosts 设置为通配符;域环境优先使用 Kerberos,跨域场景应采用受控的 HTTPS 监听器或明确的信任范围。
实现并行任务与超时回收
下面的函数把目标数组交给 Invoke-Command。PowerShell 按 ThrottleLimit 控制并发;会话选项限制连接和操作等待时间;Wait-Job -Timeout 再为整个批次设置回收边界。
Set-StrictMode -Version Latest
$ErrorActionPreference = 'Stop'
function Invoke-ParallelInspection {
[CmdletBinding()]
param(
[Parameter(Mandatory)]
[ValidateNotNullOrEmpty()]
[string[]]$ComputerName,
[ValidateRange(1, 64)]
[int]$ThrottleLimit = 5,
[ValidateRange(5, 3600)]
[int]$TimeoutSeconds = 60
)
$batchId = [guid]::NewGuid().ToString('N')
$batchStartedAt = Get-Date
$parentJob = $null
$collector = {
Set-StrictMode -Version Latest
$ErrorActionPreference = 'Stop'
$startedAt = Get-Date
$os = Get-CimInstance -ClassName Win32_OperatingSystem
$disks = @(
Get-CimInstance -ClassName Win32_LogicalDisk -Filter 'DriveType = 3' |
ForEach-Object {
$size = [double]$_.Size
$free = [double]$_.FreeSpace
[pscustomobject][ordered]@{
Name = $_.DeviceID
FreePercent = if ($size -gt 0) {
[math]::Round(($free / $size) * 100, 2)
}
else {
0
}
}
}
)
$services = @(
foreach ($name in @('WinRM', 'W32Time')) {
$service = Get-CimInstance `
-ClassName Win32_Service `
-Filter "Name = '$name'" `
-ErrorAction SilentlyContinue
[pscustomobject][ordered]@{
Name = $name
Found = $null -ne $service
State = if ($null -ne $service) { $service.State } else { $null }
}
}
)
$finishedAt = Get-Date
[pscustomobject][ordered]@{
StartedAt = $startedAt.ToString('o')
FinishedAt = $finishedAt.ToString('o')
DurationSeconds = [math]::Round(
($finishedAt - $startedAt).TotalSeconds,
2
)
OSCaption = $os.Caption
OSVersion = $os.Version
UptimeHours = [math]::Round(
($finishedAt - $os.LastBootUpTime).TotalHours,
2
)
Disks = $disks
Services = $services
}
}
function New-InspectionRow {
param(
[string]$HostName,
[string]$Status,
[AllowNull()][string]$ErrorMessage,
[AllowNull()][object]$Payload
)
$now = Get-Date
[pscustomobject][ordered]@{
BatchId = $batchId
ComputerName = $HostName
StartedAt = if ($null -ne $Payload) {
$Payload.StartedAt
}
else {
$batchStartedAt.ToString('o')
}
FinishedAt = if ($null -ne $Payload) {
$Payload.FinishedAt
}
else {
$now.ToString('o')
}
DurationSeconds = if ($null -ne $Payload) {
$Payload.DurationSeconds
}
else {
[math]::Round(($now - $batchStartedAt).TotalSeconds, 2)
}
Status = $Status
Error = $ErrorMessage
OSCaption = if ($null -ne $Payload) { $Payload.OSCaption } else { $null }
OSVersion = if ($null -ne $Payload) { $Payload.OSVersion } else { $null }
UptimeHours = if ($null -ne $Payload) { $Payload.UptimeHours } else { $null }
Disks = if ($null -ne $Payload) { @($Payload.Disks) } else { @() }
Services = if ($null -ne $Payload) { @($Payload.Services) } else { @() }
}
}
try {
$sessionOption = New-PSSessionOption `
-OpenTimeout ($TimeoutSeconds * 1000) `
-OperationTimeout ($TimeoutSeconds * 1000)
$parentJob = Invoke-Command `
-ComputerName $ComputerName `
-ScriptBlock $collector `
-SessionOption $sessionOption `
-ThrottleLimit $ThrottleLimit `
-AsJob
$null = Wait-Job -Job $parentJob -Timeout ($TimeoutSeconds + 5)
$rows = @(
foreach ($childJob in $parentJob.ChildJobs) {
$hostName = $childJob.Location
if ($childJob.State -eq 'Completed') {
$payload = @(
Receive-Job -Job $childJob -ErrorAction SilentlyContinue
) | Select-Object -First 1
if ($null -ne $payload) {
New-InspectionRow `
-HostName $hostName `
-Status 'Success' `
-ErrorMessage $null `
-Payload $payload
}
else {
New-InspectionRow `
-HostName $hostName `
-Status 'Failed' `
-ErrorMessage 'Remote job completed without a payload.' `
-Payload $null
}
}
elseif ($childJob.State -in @('Running', 'NotStarted')) {
Stop-Job -Job $childJob -ErrorAction SilentlyContinue
New-InspectionRow `
-HostName $hostName `
-Status 'Timeout' `
-ErrorMessage "Inspection exceeded $TimeoutSeconds seconds." `
-Payload $null
}
else {
$reason = $childJob.JobStateInfo.Reason
$message = if ($null -ne $reason) {
$reason.Message
}
else {
"Remote job ended in state '$($childJob.State)'."
}
New-InspectionRow `
-HostName $hostName `
-Status 'Failed' `
-ErrorMessage $message `
-Payload $null
}
}
)
# 即使连接初始化未形成子任务,也为输入主机补一条失败记录。
$knownHosts = @($rows | ForEach-Object { $_.ComputerName })
foreach ($hostName in $ComputerName) {
if ($hostName -notin $knownHosts) {
$rows += New-InspectionRow `
-HostName $hostName `
-Status 'Failed' `
-ErrorMessage 'No child job was created for this host.' `
-Payload $null
}
}
$rows | Sort-Object -Property ComputerName
}
catch {
foreach ($hostName in $ComputerName) {
New-InspectionRow `
-HostName $hostName `
-Status 'Failed' `
-ErrorMessage $_.Exception.Message `
-Payload $null
}
}
finally {
if ($null -ne $parentJob) {
Remove-Job -Job $parentJob -Force -ErrorAction SilentlyContinue
}
}
}
# 先用本机验证 Remoting,再替换为真实服务器数组。
$targets = @($env:COMPUTERNAME)
$inspection = @(
Invoke-ParallelInspection `
-ComputerName $targets `
-ThrottleLimit 5 `
-TimeoutSeconds 30
)
ConvertTo-Json -InputObject $inspection -Depth 6预期输出如下。BatchId、主机名、时间和指标值均会随运行环境变化。
[
{
"BatchId": "<ID>",
"ComputerName": "<HOST>",
"StartedAt": "<TIME>",
"FinishedAt": "<TIME>",
"DurationSeconds": 0.42,
"Status": "Success",
"Error": null,
"OSCaption": "Microsoft Windows Server 2022 Standard",
"OSVersion": "10.0.20348",
"UptimeHours": 168.25,
"Disks": [
{
"Name": "C:",
"FreePercent": 61.37
}
],
"Services": [
{
"Name": "WinRM",
"Found": true,
"State": "Running"
},
{
"Name": "W32Time",
"Found": true,
"State": "Running"
}
]
}
]验证节流、超时与关联
先用一台已配置 Remoting 的主机验证字段,再逐步扩大目标数组。把 ThrottleLimit 设为较小值可以观察任务分批运行;加入一台不可达主机后,应得到 Failed 或 Timeout,且记录仍包含原始 ComputerName 和相同 BatchId。
远程任务完成顺序不保证与输入顺序一致,不能依赖数组位置关联主机。脚本按主机名排序只是为了阅读,真正的关联键是 BatchId + ComputerName。写入数据库时,可再增加规则版本或任务编号形成唯一键。
生产边界与故障定位
ThrottleLimit 不是越大越快。并发过高会同时消耗发起端运行空间、网络连接、目标端 WinRM 配额和 CIM 资源。应从小值开始,根据平均耗时、失败率和目标端负载逐步调整。
大量 Failed 应先检查 DNS、Kerberos、WinRM 监听器和防火墙;只有部分主机 Timeout 时,再检查目标负载、查询范围和超时阈值。不要通过关闭证书校验、扩大 TrustedHosts 或长期使用高权限账号掩盖配置问题。