#Requires -Version 5.1
<#
    JbTecWiz Support Centre -- generated fix script

    Fault : The cluster will not start -- quorum lost, event 1177
    Fix   : Force quorum to bring the cluster back
    Source: https://jbtecwiz.com/support/srv-clu-quorum

    Run as     : Elevated PowerShell on the surviving node
    Expect     : 40 minutes
    Risk       : high
    Reversible : yes

    WHEN THIS IS THE RIGHT FIX
    Enough nodes are genuinely lost that quorum cannot be reached. Read the
    warning first.

    HOW TO UNDO IT
    Weights can be reset. The forced start itself cannot be undone, which is
    why the confirmation in step one matters.

    This script walks the fix one step at a time and asks before each
    one. Steps with no command are things you do yourself -- it prints
    those and waits. Run with -DryRun to print without executing.

    --------------------------------------------------------------------
    NO WARRANTY - USE AT YOUR OWN RISK

    This script is provided by JbTecWiz as-is and with no warranty of any
    kind, express or implied. You run it entirely at your own risk.

    JbTecWiz accepts no liability for any loss or damage arising from its
    use, including but not limited to data loss, downtime, or configuration
    changes that turn out to be wrong for your system.

    You are responsible for reading this script before running it, for
    satisfying yourself that it suits the machine in front of you, and for
    having a working backup first. Some steps cannot be undone.
    --------------------------------------------------------------------
#>

[CmdletBinding()]
param(
    # Print every step and command without running anything.
    [switch]$DryRun,
    # Do not ask before each step. Read the script first if you use this.
    [switch]$Unattended
)

$ErrorActionPreference = 'Stop'
$script:Failed = 0

function Write-Rule { param([string]$Text)
    Write-Host ''
    Write-Host ('-' * 70) -ForegroundColor DarkGray
    if ($Text) { Write-Host $Text -ForegroundColor Cyan }
}

function Show-Prose { param([string]$Text, [string]$Colour = "Gray")
    if (-not $Text) { return }
    $words = $Text -split "\s+"; $line = "  "
    foreach ($w in $words) {
        if (($line.Length + $w.Length) -gt 74) { Write-Host $line -ForegroundColor $Colour; $line = "  " }
        $line += "$w "
    }
    if ($line.Trim()) { Write-Host $line -ForegroundColor $Colour }
}

function Invoke-Step {
    param(
        [int]$Number, [string]$Do, [string]$Why,
        [scriptblock]$Command, [switch]$Manual, [string]$Shell = "powershell"
    )
    Write-Rule "  Step $Number of 6"
    Show-Prose $Do "White"
    if ($Why) { Write-Host ""; Show-Prose $Why "DarkGray" }

    if ($Manual) {
        Write-Host ''
        Write-Host '  -> Do this yourself, then press Enter to carry on.' -ForegroundColor Yellow
        if (-not $Unattended -and -not $DryRun) { [void](Read-Host) }
        return
    }

    Write-Host ''
    foreach ($l in ($Command.ToString().Trim() -split "`n")) {
        Write-Host ("    " + $l.Trim()) -ForegroundColor Green
    }
    Write-Host ''

    if ($DryRun) { Write-Host "  (dry run -- not executed)" -ForegroundColor DarkGray; return }
    if (-not $Unattended) {
        $a = Read-Host "  Run this step? [Y]es / [S]kip / [Q]uit"
        if ($a -match "^[Qq]") { Write-Host "  Stopped at your request."; exit 0 }
        if ($a -match "^[Ss]") { Write-Host "  Skipped." -ForegroundColor DarkGray; return }
    }
    try { & $Command }
    catch {
        $script:Failed++
        Write-Host ("  Step $Number failed: " + $_.Exception.Message) -ForegroundColor Red
        Show-Prose "The rest of the fix may depend on this. Read the write-up before carrying on." "Red"
        if (-not $Unattended) {
            $c = Read-Host "  Carry on anyway? [y/N]"
            if ($c -notmatch "^[Yy]") { exit 1 }
        }
    }
}

Write-Rule
Write-Host '  The cluster will not start -- quorum lost, event 1177' -ForegroundColor White
Write-Host '  Force quorum to bring the cluster back' -ForegroundColor Cyan
Write-Host ''
Write-Host '  Risk: high   Reversible   40 minutes'
Write-Host ''
Show-Prose 'No warranty. Use at your own risk - JbTecWiz accepts no liability. Read it before you run it, and have a backup.' 'DarkYellow'
Write-Rule

if (-not $Unattended -and -not $DryRun) {
    $go = Read-Host '  Ready? [y/N]'
    if ($go -notmatch "^[Yy]") { Write-Host "  Nothing was changed."; exit 0 }
}

Invoke-Step -Number 1 -Do 'Be certain the other nodes are genuinely down and not merely unreachable. Forcing quorum while another partition is also running produces two clusters writing to the same storage.' -Why 'This is the split-brain scenario the quorum mechanism exists to prevent, and it is the one way to make a bad outage into data loss. Confirm the other nodes are powered off, not just unreachable.' -Manual

Invoke-Step -Number 2 -Do 'Confirm the state of every node before forcing anything.' -Command {
    Get-ClusterNode | Format-Table Name,State
    Test-NetConnection node02 -Port 3343
}

Invoke-Step -Number 3 -Do 'Start the cluster service with quorum forced on the surviving node.' -Command {
    Start-ClusterNode -Name node01 -FixQuorum
}

Invoke-Step -Number 4 -Do 'Give the surviving node full weight so the cluster is stable while the others are down.' -Command {
    (Get-ClusterNode node01).NodeWeight = 1
    Get-ClusterNode | Where-Object Name -ne 'node01' | ForEach-Object { $_.NodeWeight = 0 }
}

Invoke-Step -Number 5 -Do 'Bring the other nodes back one at a time, restoring their weight as each joins.' -Command {
    Start-ClusterNode -Name node02
    (Get-ClusterNode node02).NodeWeight = 1
}

Invoke-Step -Number 6 -Do 'Restore the normal quorum configuration once every node is back.' -Command {
    Get-ClusterNode | ForEach-Object { $_.NodeWeight = 1 }
    Get-ClusterQuorum
}

Write-Rule "  Confirm it worked"
Show-Prose 'All nodes are Up with equal weight and the cluster validates.' 'White'
Write-Host ''
if (-not $DryRun) {
    Get-ClusterNode | Format-Table Name,State,NodeWeight
    Test-Cluster -Node (Get-ClusterNode).Name -Include 'Inventory','Network','System Configuration'
}

Write-Rule
if ($script:Failed -gt 0) {
    Write-Host ("  Finished with " + $script:Failed + " failed step(s).") -ForegroundColor Yellow
    Show-Prose 'Read the full write-up at https://jbtecwiz.com/support/srv-clu-quorum' 'Yellow'
} else {
    Write-Host '  Finished.' -ForegroundColor Green
}
Write-Host ''
Show-Prose 'To undo: Weights can be reset. The forced start itself cannot be undone, which is why the confirmation in step one matters.' 'DarkGray'
Write-Rule