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

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

    Run as     : Elevated PowerShell
    Expect     : 60 minutes
    Risk       : medium
    Reversible : yes

    WHEN THIS IS THE RIGHT FIX
    Nodes are running but cannot see each other.

    HOW TO UNDO IT
    Cluster network roles can be set back with Set-ClusterNetwork.

    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 '  Repair the cluster network' -ForegroundColor Cyan
Write-Host ''
Write-Host '  Risk: medium   Reversible   60 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 'Look at the cluster networks and their roles.' -Command {
    Get-ClusterNetwork | Format-Table Name,State,Role,Address,AddressMask
    Get-ClusterNetworkInterface | Format-Table Name,Node,Network,State
}

Invoke-Step -Number 2 -Do 'Test the cluster heartbeat port between nodes.' -Why 'Cluster communication is UDP 3343 with a TCP fallback. A firewall change or a VLAN alteration that blocks it partitions the cluster while every other kind of traffic keeps working, which makes it look like a cluster fault rather than a network one.' -Command {
    Test-NetConnection node02 -Port 3343 -InformationLevel Detailed
}

Invoke-Step -Number 3 -Do 'Check the heartbeat thresholds -- on a busy or virtualised cluster the defaults can be too tight.' -Command {
    Get-Cluster | Format-List SameSubnetDelay,SameSubnetThreshold,CrossSubnetDelay,CrossSubnetThreshold
}

Invoke-Step -Number 4 -Do 'Check for a network adapter that has been reconfigured, or a teaming change that dropped the cluster network.' -Command {
    Get-NetAdapter | Format-Table Name,Status,LinkSpeed,InterfaceDescription
    Get-ClusterNetwork | Where-Object State -ne 'Up' | Format-List
}

Invoke-Step -Number 5 -Do 'Run the validation report for the network only -- it is quick and does not disrupt running roles.' -Command {
    Test-Cluster -Node node01,node02 -Include 'Network' -ReportName C:\clusternet
}

Invoke-Step -Number 6 -Do 'Check the cluster log for the moment of the partition.' -Command {
    Get-ClusterLog -TimeSpan 30 -Destination C:\clusterlogs
}

Write-Rule "  Confirm it worked"
Show-Prose 'Every cluster network is Up and the validation report has no network failures.' 'White'
Write-Host ''
if (-not $DryRun) {
    Get-ClusterNetwork | Format-Table Name,State,Role
}

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: Cluster network roles can be set back with Set-ClusterNetwork.' 'DarkGray'
Write-Rule