Runaway process checking

Recently I ran into an issue with PHP exhaustion on a Windows Server running IIS. In this scenario the PHP-CGI.exe process would continue to spawn additional instances as load on the server would increase but over time the application pool would struggle and begin to slow to a crawl. In the past I have seen other applications during various iterations of development run into the same issue where if you run into more than “x” instances of an application it is unhealthy or less than “y” instances it is not running properly.

 

$myprocess = "php-cgi"
$myserver = "WebServer"
$mydomain = "liquidobject.com"
$mail_server = "mail.liquidobject.com"
$mail_recipient = "my_support_team@liquidobject.com"
$toomany = 40
$waytoomany = 80

$mail_sender = "$myserver@$mydomain"
$mailreport_subject = "Script: $myserver $myprocess count"
$body = " "

function SendEmailReport
{
    $body = [string]::join([environment]::NewLine, ($body)) 
    $msg = New-Object System.Net.Mail.MailMessage $mail_sender, $mail_recipient, $mailreport_subject, $body
    $client = New-Object System.Net.Mail.SmtpClient $mail_server
    $client.Credentials = [System.Net.CredentialCache]::DefaultNetworkCredentials
    $client.Send($msg)
}


$mycount =  (Get-Process -Name $myprocess).count

if($toomany -lt $mycount)
{
	$body = "We have $mycount $myprocess processes, something is unusual."
	if($waytoomany -lt $mycount)
        {
             IISRESET /STOP
             IISRESET /START
             $body = "We have $mycount $myprocess processes, IIS has been reset."
        }
        SendEmailReport
}

In this case we are sending an email notification to the fictional “support team” when more than 40 instances of the php-cgi process are running and in the event no one responds by the time 80 instances are hit the site is automatically bounced to ensure it’s availability.

The simple method for checking is the use of Task Scheduler and call up the script every 5 minutes, pretty simple yet effective.