ROMFASTSQL/proxmox/cluster/scripts/wake-cluster.ps1

<#
.SYNOPSIS
    Porneste nodurile Proxmox prin Wake-on-LAN.

.DESCRIPTION
    Trimite magic packet catre placile de retea de PRODUCTIE ale nodurilor si
    asteapta sa apara pe retea.

    De ce merge abia din 2026-08-29: pana atunci pve1 si pveelite aveau IP-ul
    principal pe dongle-uri USB Realtek, care nu suporta WoL si oricum pierd
    alimentarea cand nodul e oprit. Mutarea lui vmbr0 pe interfetele onboard a
    facut WoL posibil - vezi ../docs/switch-2.5g-dedicat-noduri.md.

    NU trezi un singur nod si atat, daca ai oprit tot clusterul: un nod pornit
    singur nu are quorum, /etc/pve ramane read-only si nu poti porni niciun guest.
    Implicit scriptul le trezeste pe toate trei.

    RULEAZA DE PE STATIA DE ADMIN, care trebuie sa fie pe acelasi segment L2
    (10.0.20.0/24) - magic packet-ul e broadcast si nu trece prin router.

.PARAMETER Node
    Trezeste doar nodul dat. Implicit: toate trei.

.PARAMETER Wait
    Secunde de asteptat sa apara nodurile (implicit 300).

.PARAMETER NoWait
    Trimite pachetele si iese, fara sa astepte.

.EXAMPLE
    .\wake-cluster.ps1
.EXAMPLE
    .\wake-cluster.ps1 -Node pveelite
#>
[CmdletBinding()]
param(
    [ValidateSet('pve1','pvemini','pveelite')]
    [string]$Node,
    [int]$Wait = 300,
    [switch]$NoWait
)

$ErrorActionPreference = 'Stop'

# MAC-ul interfetei de PRODUCTIE (portul puntii vmbr0), nu al dongle-ului de insula.
# Citite de pe noduri pe 2026-08-29 cu: ethtool <if>; cat /sys/class/net/<if>/address
$Nodes = [ordered]@{
    pve1     = @{ Ip = '10.0.20.200'; Mac = 'fc:3f:db:0a:0d:d8'; If = 'eno1'    }
    pvemini  = @{ Ip = '10.0.20.201'; Mac = '58:47:ca:7d:51:3c'; If = 'enp90s0' }
    pveelite = @{ Ip = '10.0.20.202'; Mac = '84:69:93:57:b2:ea'; If = 'eno1'    }
}

function Write-Step { param([string]$Text) Write-Host ""; Write-Host "== $Text" -ForegroundColor Green }
function Write-Ok   { param([string]$Text) Write-Host "  [ok] $Text" -ForegroundColor Green }
function Write-Warn { param([string]$Text) Write-Host "  [!]  $Text" -ForegroundColor Yellow }
function Write-Info { param([string]$Text) Write-Host "       $Text" -ForegroundColor DarkGray }

function Test-Alive {
    param([string]$Address, [int]$TimeoutMs = 1500)
    try {
        $p = New-Object System.Net.NetworkInformation.Ping
        return ($p.Send($Address, $TimeoutMs).Status -eq 'Success')
    } catch { return $false }
}

function Send-MagicPacket {
    param([string]$Mac)

    $bytes = ($Mac -split '[:-]') | ForEach-Object { [byte]([Convert]::ToInt32($_, 16)) }
    if ($bytes.Count -ne 6) { throw "MAC invalid: $Mac" }

    # Magic packet: 6 octeti 0xFF, apoi MAC-ul repetat de 16 ori.
    $packet = New-Object byte[] 102
    for ($i = 0; $i -lt 6; $i++) { $packet[$i] = 0xFF }
    for ($r = 0; $r -lt 16; $r++) { [Array]::Copy($bytes, 0, $packet, 6 + $r * 6, 6) }

    # Trimis si pe broadcast global, si pe cel de subretea: unele stive le trateaza
    # diferit, iar pachetul e mic - nu costa nimic sa mearga pe ambele.
    # Porturile 9 si 7 sunt cele uzuale; placile asculta oricum dupa continut.
    $udp = New-Object System.Net.Sockets.UdpClient
    try {
        $udp.EnableBroadcast = $true
        foreach ($dest in @('255.255.255.255', '10.0.20.255')) {
            foreach ($port in @(9, 7)) {
                [void]$udp.Send($packet, $packet.Length, $dest, $port)
            }
        }
    } finally {
        $udp.Close()
    }
}

$targets = if ($Node) { @($Node) } else { $Nodes.Keys }

Write-Step "Trimit magic packet"

$deja = @()
foreach ($n in $targets) {
    if (Test-Alive $Nodes[$n].Ip) {
        Write-Ok "$n ($($Nodes[$n].Ip)) - deja pornit, nu trimit"
        $deja += $n
        continue
    }
    Send-MagicPacket -Mac $Nodes[$n].Mac
    Write-Ok "$n -> $($Nodes[$n].Mac) ($($Nodes[$n].If))"
}

$asteptate = @($targets | Where-Object { $deja -notcontains $_ })

if ($asteptate.Count -eq 0) {
    Write-Warn "Toate nodurile cerute erau deja pornite."
    exit 0
}

if ($NoWait) {
    Write-Info "-NoWait: nu astept. Verifica singur cu ping."
    exit 0
}

Write-Step "Astept nodurile (max ${Wait}s)"

# Retrimitem periodic: daca placa nu era inca alimentata la primul pachet
# (alimentare tocmai revenita, de exemplu), al doilea o prinde.
$waited = 0
$up = @()
while ($waited -lt $Wait) {
    foreach ($n in $asteptate) {
        if ($up -contains $n) { continue }
        if (Test-Alive $Nodes[$n].Ip) {
            $up += $n
            Write-Ok "$n a pornit (${waited}s)"
        }
    }
    if ($up.Count -eq $asteptate.Count) { break }

    Start-Sleep -Seconds 5
    $waited += 5

    if ($waited % 60 -eq 0) {
        foreach ($n in $asteptate) {
            if ($up -notcontains $n) { Send-MagicPacket -Mac $Nodes[$n].Mac }
        }
        Write-Info "am retrimis pachetele (${waited}s)"
    }
}

$lipsa = @($asteptate | Where-Object { $up -notcontains $_ })

Write-Step "Rezultat"

if ($lipsa.Count -eq 0) {
    Write-Ok "toate nodurile cerute au pornit"
    Write-Host ""
    Write-Host "  Urmatorul pas: .\cluster-startup.ps1"
    Write-Host ""
} else {
    Write-Warn "NU au pornit: $($lipsa -join ', ')"
    Write-Info "Cauza cea mai probabila e in BIOS, nu in retea:"
    Write-Info "  - 'Wake on LAN' / 'Power On by PCI-E' dezactivat"
    Write-Info "  - ErP Ready / Deep Sleep ACTIVAT (taie alimentarea placii in S5)"
    Write-Info "Verifica si ca statia de admin e pe acelasi segment L2 (10.0.20.0/24)."
    Write-Info "Pe nodurile pornite, starea WoL se vede cu: ethtool <if> | grep Wake-on"
    exit 1
}