#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   : Find the scheduled load that is starving the heartbeat
    Source: https://jbtecwiz.com/support/srv-cluster-1135

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

    WHEN THIS IS THE RIGHT FIX
    Evictions cluster around a particular time.

    HOW TO UNDO IT
    None -- this is scheduling and topology, not configuration to undo.

    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 5"
    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 '  Find the scheduled load that is starving the heartbeat' -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 'Note the exact times from the event log and look for what else runs then.' -Command {
    Get-WinEvent -FilterHashtable @{LogName='System'; Id=1135} -MaxEvents 30 | Select-Object TimeCreated
}

Invoke-Step -Number 2 -Do 'Check backup schedules first -- a backup reading a CSV at full speed is the classic cause.' -Manual

Invoke-Step -Number 3 -Do 'Check for scheduled antivirus full scans.' -Command {
    Get-MpPreference | Select-Object ScanScheduleDay, ScanScheduleTime
}

Invoke-Step -Number 4 -Do 'Separate the heartbeat onto its own physical network if it currently shares with storage or backup traffic.' -Manual

Invoke-Step -Number 5 -Do 'If the nodes are virtual, check the host for the same window -- host-level backups that quiesce a guest will pause it past the heartbeat threshold.' -Manual

Write-Rule "  Confirm it worked"
Show-Prose 'The window passes without an eviction.' 'White'

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: None -- this is scheduling and topology, not configuration to undo.' 'DarkGray'
Write-Rule