Monthly Archives: February 2013

Finding SQL Server Installs Using Powershell

Old Dog, New Tricks

I’ve been writing tools for SQL Server for a lot of years. Some of these tools were never completely released. Some of them were just for me. Some of them overlapped other tools already on the market and free for all. Recently, I started updating my bag of tricks and tools. This seemed like a great time to get back into PowerShell. I decided to pull out a bit of C# code I cobbled together nine years ago as part of a tool to find SQL Server instances on a network. I never really got around to making it a “production” ready tool since there was already a most excellent one on the scene in the form of SQLPing from Chip Andrews. SQLPing is a fantastic scan tool and can scan many more things than the method covered here.

Hello Operator?

When Microsoft implemented named instances with SQL Server 2000 they had to have a way to direct incoming traffic pointed to a single IP to the correct TCP port. So, they built in a traffic cop. When SQL Server 2005 came around it was pulled from the core network listener and put into its own service, the browser service. There was little documentation on how the browser worked. Luckily for poor sods like me using a network packet sniffing tool it was pretty easy to figure out how to talk to the browser. Once I figured out how to get the browser service to tell me what instances it knew about it was trivial to implement. These days Microsoft is being much more open about these kinds of things and actually have released documentation on how the browser service and the SQL Server Resolution Protocol works.

The Basic Mechanics.

As most of you know SQL Server’s default instance is on 1433 and the browser service is on 1434. Our goal is to send a UDP packet to port 1434. According to the docs we only need to send a single byte containing the number two. This prompts the listener to give us a list of instances and what port they are bound to. When I wrote my implementation it really was that simple. I dug around and figured out how to get PowerShell to send a UDP packet. I tested it and lo’ it worked, on my machine….

It’s Never So Easy.

When I tested it on my lab VM cluster with multiple nodes and multiple instances it would fail! I just didn’t get it. My C# code from the stone age worked just fine. My PowerShell code was a hit or miss. I started troubleshooting the problem just as I had in the beginning. I fired up my network sniffer and watched the traffic flow back and fourth. Again, I saw exactly what I expected, or more accurately what I wanted to see.  A single packet with the hex number two in the first byte position. I ran the test several times over the next hour or so. Eventually, I just had to walk away from it. The next day I started over again. I read the documentation, it still said the same thing. I ran the test, still a two in the first byte position. Then I spotted it. The packet wasn’t one byte long. I went back and read the document again. It gives an upper boundary of 32 bytes but no lower limit. The packet I saw come through wasn’t one byte long it was always more than that. Armed with that I started big and worked my way down until I got errors. Now I know that a packet of three bytes always triggers a response. a  two with two zeros.

Meet QuerySQLListener.

Here is the function I put together. It takes a server name and returns an array of strings with the fun bits in it.

function QuerySQLListener{
    [cmdletbinding(
        DefaultParameterSetName = '',
        ConfirmImpact = "low"
    )]
    Param(
        [Parameter(
            Mandatory = $True,
            Position = 0,
            ParameterSetName = '',
            ValueFromPipeline = $True)]
        [string]$Computer
    )
    Begin {
        $ErrorActionPreference = "SilentlyContinue"
        $Port = 1434
        $ConnectionTimeout = 1000
        $Responses  = @();
    }
    Process {
        $UDPClient = new-Object system.Net.Sockets.Udpclient
        $UDPClient.client.ReceiveTimeout = $ConnectionTimeout
        $IPAddress = [System.Net.Dns]::GetHostEntry($Computer).AddressList[0].IPAddressToString
        $UDPClient.Connect($IPAddress,$Port)
        $ToASCII = new-object system.text.asciiencoding
        $UDPPacket = 0x02,0x00,0x00
        Try {
            $UDPEndpoint = New-Object system.net.ipendpoint([system.net.ipaddress]::Any,0)
            $UDPClient.Client.Blocking = $True
            [void]$UDPClient.Send($UDPPacket,UDPPacket.length)
            $BytesRecived = $UDPClient.Receive([ref]$UDPEndpoint)
            [string]$Response = $ToASCII.GetString($BytesRecived)
            $res = ""
            If ($Response) {
                $Response = $Response.Substring(3,$Response.Length-3).Replace(";;","~")
                #$i = 0;
                $Response.Split("~") | ForEach {
                $Responses += $_
            }
            $socket = $null;
            $UDPClient.close()
        }
        }
        Catch {
            $Error[0].ToString()
            $UDPClient.Close()
        }
    }
    End {
        return ,$Responses
    }
}

 

It Isn’t Perfect But It Works.

I”m sure there is a cleaner way to implement it but I’m really just getting into PowerShell again after several months of tinkering with it last time. If you have any suggestions or improvements I’ll gladly take them!