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

    Fault : Failover Cluster event 1135 -- a node was removed from the active failover cluster membership
    Fix   : Rule out the network and the NIC settings
    Source: https://jbtecwiz.com/support/srv-cluster-1135

    Run as     : Elevated PowerShell on each node
    Expect     : 1 hour
    Risk       : low
    Reversible : yes

    WHEN THIS IS THE RIGHT FIX
    No obvious time pattern.

    HOW TO UNDO IT
    Return SameSubnetDelay to 1000 and SameSubnetThreshold to 5 (the
    defaults for recent versions) if you changed them.

    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 '  Failover Cluster event 1135 -- a node was removed from the active failover cluster membership' -ForegroundColor White
Write-Host '  Rule out the network and the NIC settings' -ForegroundColor Cyan
Write-Host ''
Write-Host '  Risk: low   Reversible   1 hour'
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 'Collect the cluster log around the eviction -- it records the heartbeat misses in detail.' -Command {
    Get-ClusterLog -TimeSpan 15 -Destination C:\temp
}

Invoke-Step -Number 2 -Do 'Turn off power management on every cluster NIC.' -Why 'A NIC allowed to sleep will drop heartbeats on an otherwise idle network, which produces exactly this event at unpredictable times.' -Command {
    Get-NetAdapter | Get-NetAdapterPowerManagement | Where-Object AllowComputerToTurnOffDevice -eq 'Enabled'
}

Invoke-Step -Number 3 -Do 'Disable it where enabled.' -Command {
    Set-NetAdapterPowerManagement -Name 'Cluster' -AllowComputerToTurnOffDevice Disabled
}

Invoke-Step -Number 4 -Do 'Exclude cluster traffic and the cluster directories from antivirus on-access scanning.' -Manual

Invoke-Step -Number 5 -Do 'Check the cluster network configuration is sane -- at least one network dedicated to cluster communication.' -Command {
    Get-ClusterNetwork | Format-Table Name, Role, Address, State -AutoSize
}

Invoke-Step -Number 6 -Do 'Only if the physical network is known-good and cannot be improved, relax the heartbeat thresholds.' -Why 'This is a last resort. It makes the cluster slower to detect a genuine node failure -- it hides the symptom rather than fixing the cause.' -Command {
    (Get-Cluster).SameSubnetDelay = 1000
    (Get-Cluster).SameSubnetThreshold = 10
}

Write-Rule "  Confirm it worked"
Show-Prose 'No further 1135 events over a full week.' 'White'
Write-Host ''
if (-not $DryRun) {
    Get-WinEvent -FilterHashtable @{LogName='System'; Id=1135} -MaxEvents 20 | Format-Table TimeCreated, Message -AutoSize
}

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-cluster-1135' 'Yellow'
} else {
    Write-Host '  Finished.' -ForegroundColor Green
}
Write-Host ''
Show-Prose 'To undo: Return SameSubnetDelay to 1000 and SameSubnetThreshold to 5 (the defaults for recent versions) if you changed them.' 'DarkGray'
Write-Rule