local-ci-cd-system/scripts/Invoke-RetentionPolicy.ps1

#Requires -Version 5.1
<#
.SYNOPSIS
    Purges old CI artifacts, logs, and stale IP leases based on retention policy.

.DESCRIPTION
    Removes per-job subdirectories under ArtifactDir and LogDir whose last-write
    time is older than RetentionDays. If the monitored drive is below MinFreeGB,
    switches automatically to AggressiveRetentionDays to reclaim space faster.

    Also removes stale IP lease files under F:\CI\State\ip-leases\ that are
    older than 12 hours (defensive cleanup for jobs that crashed without a finally).

    Designed to run as a Windows Scheduled Task (see Register-CIScheduledTasks.ps1).
    Safe to run while CI jobs are active — only deletes directories whose
    last-write time predates the cutoff.

.PARAMETER ArtifactDir
    Directory containing per-job artifact subdirectories. Default: F:\CI\Artifacts

.PARAMETER LogDir
    Directory containing per-job log subdirectories. Default: F:\CI\Logs

.PARAMETER RetentionDays
    Normal retention threshold in days. Default: 30

.PARAMETER AggressiveRetentionDays
    Retention threshold used when drive free space drops below MinFreeGB. Default: 7

.PARAMETER MinFreeGB
    If drive free space is below this value, AggressiveRetentionDays is used instead
    of RetentionDays and a warning is emitted. Default: 50

.PARAMETER DriveLetter
    Single drive letter (without colon) to check for free space. Default: F

.EXAMPLE
    # Dry run — lists what would be deleted
    .\Invoke-RetentionPolicy.ps1 -WhatIf

    # Live run
    .\Invoke-RetentionPolicy.ps1
#>
[CmdletBinding(SupportsShouldProcess)]
param(
    [string] $ArtifactDir = 'F:\CI\Artifacts',
    [string] $LogDir      = 'F:\CI\Logs',

    [ValidateRange(1, 365)]
    [int] $RetentionDays = 30,

    [ValidateRange(1, 365)]
    [int] $AggressiveRetentionDays = 7,

    [ValidateRange(1, 1000)]
    [int] $MinFreeGB = 50,

    [ValidatePattern('^[A-Za-z]$')]
    [string] $DriveLetter = 'F'
)

Set-StrictMode -Version Latest
$ErrorActionPreference = 'Continue'   # Don't abort on individual file errors

# ── Determine effective retention threshold ───────────────────────────────────
$drive = Get-PSDrive -Name $DriveLetter -ErrorAction SilentlyContinue
$freeGB = if ($drive) { [math]::Round($drive.Free / 1GB, 1) } else { [double]::MaxValue }

$aggressiveMode  = $freeGB -lt $MinFreeGB
$effectiveDays   = if ($aggressiveMode) { $AggressiveRetentionDays } else { $RetentionDays }
$cutoff          = (Get-Date).AddDays(-$effectiveDays)

Write-Host "[RetentionPolicy] Drive ${DriveLetter}: free ${freeGB} GB"
if ($aggressiveMode) {
    Write-Warning "[RetentionPolicy] Free space below ${MinFreeGB} GB — aggressive retention: ${AggressiveRetentionDays}d"
} else {
    Write-Host "[RetentionPolicy] Normal retention: ${effectiveDays}d (cutoff: $($cutoff.ToString('yyyy-MM-dd')))"
}

# ── Helper: purge old subdirectories under a base dir ─────────────────────────
function Remove-OldJobDirs {
    [CmdletBinding(SupportsShouldProcess)]
    param([string]$BaseDir, [string]$Label)

    if (-not (Test-Path $BaseDir)) {
        Write-Host "[RetentionPolicy] $Label dir not found: $BaseDir — skipping."
        return
    }

    $old = Get-ChildItem -Path $BaseDir -Directory -ErrorAction SilentlyContinue |
        Where-Object { $_.LastWriteTime -lt $cutoff }

    if (-not @($old).Count) {
        Write-Host "[RetentionPolicy] ${Label}: nothing to purge (cutoff: $($cutoff.ToString('yyyy-MM-dd')))."
        return
    }

    foreach ($dir in $old) {
        $agedays = [int]((Get-Date) - $dir.LastWriteTime).TotalDays
        if ($PSCmdlet.ShouldProcess($dir.FullName, "Remove $Label dir (age: ${agedays}d)")) {
            Remove-Item $dir.FullName -Recurse -Force -ErrorAction SilentlyContinue
            Write-Host "[RetentionPolicy] Purged $Label (${agedays}d): $($dir.Name)"
        }
    }
}

Remove-OldJobDirs -BaseDir $ArtifactDir -Label 'artifact'
Remove-OldJobDirs -BaseDir $LogDir      -Label 'log'

# ── Stale IP lease cleanup (§2.1 defensive) ───────────────────────────────────
# Lease files are normally deleted by Invoke-CIJob.ps1 finally block.
# If the host crashed mid-job the lease file is never removed. Remove leases
# older than 12 hours (well beyond any plausible build duration).
$leasesDir    = 'F:\CI\State\ip-leases'
$leaseCutoff  = (Get-Date).AddHours(-12)
if (Test-Path $leasesDir) {
    $staleLeases = Get-ChildItem -Path $leasesDir -File -Filter '*.lease' -ErrorAction SilentlyContinue |
        Where-Object { $_.LastWriteTime -lt $leaseCutoff }
    foreach ($lf in @($staleLeases)) {
        if ($PSCmdlet.ShouldProcess($lf.FullName, 'Remove stale IP lease')) {
            Remove-Item $lf.FullName -Force -ErrorAction SilentlyContinue
            Write-Host "[RetentionPolicy] Removed stale IP lease: $($lf.Name)"
        }
    }
}

# ── Stale vm-start.lock cleanup ─────────────────────────────────────────────
# The lock is released by Invoke-CIJob.ps1 normally. On host crash it persists;
# the OS-level lock is gone after reboot but the file blocks subsequent IP-acquire
# retries for 10 minutes and confuses operators. Remove if older than 30 minutes.
$lockFile  = 'F:\CI\State\vm-start.lock'
$lockCutoff = (Get-Date).AddMinutes(-30)
if ((Test-Path $lockFile) -and (Get-Item $lockFile).LastWriteTime -lt $lockCutoff) {
    if ($PSCmdlet.ShouldProcess($lockFile, 'Remove stale vm-start.lock')) {
        $lockAgeMin = [int]((Get-Date) - (Get-Item $lockFile).LastWriteTime).TotalMinutes
        Remove-Item $lockFile -Force -ErrorAction SilentlyContinue
        Write-Host "[RetentionPolicy] Removed stale vm-start.lock (age: ${lockAgeMin} min)"
    }
}

# ── Summary ───────────────────────────────────────────────────────────────────
$freeGBAfter = if ($drive) { [math]::Round((Get-PSDrive -Name $DriveLetter -ErrorAction SilentlyContinue).Free / 1GB, 1) } else { $freeGB }
$delta       = [math]::Round($freeGBAfter - $freeGB, 1)
Write-Host "[RetentionPolicy] Done. Drive ${DriveLetter}: free ${freeGBAfter} GB (delta: +${delta} GB)"