Designing a Hybrid Image

For some time, thin images were all the rage, IT Pros were trying to keep images as thin as possible and let the applications get installed either as part of the next task sequence or from group policy. This was in response to the old school and monolithic images that had become prevalent over the prior decade. In the past, you’d just create one thick image, per model and roll that out with everything, regardless of whether the user needed all these applications or not.

I assert however that like most things in this world, the best solution fits somewhere close to the middle of two extremes. Organizations need a balance of time and of disk space. This solution is colloquially known as a hybrid image. It allows us to keep common applications in the image and then let the supplemental applications come only later to those who need them.

First things first; don’t build your images on physical hardware. I can’t tell you how many times I still get emails from people who do this. If you are building images on physical hardware like it’s 1999, you’re doing it wrong. Always, always, always build your image in a VM and capture it via standard client task sequences. This ensures your image runs on everything, laptops, desktops, tablets. Let MDT install drivers for you. It’s that whole rule of working smarter not harder at hand here.

Next, build a hybrid image. Essentials like Office, Adobe Reader, and Some Core Line of Business Applications should go in your image. Leave Flash and Java and other commonly updated optional apps out of the image. The whole point of imaging is saving time, and while yes you can get MDT to drop a raw windows wim file and do the whole office and patching for you, this is not something you want to do repeatedly. You don’t need five dozen some PCs all pulling six month old office patches across a 40 meg connection across your WAN just because you can.

Fundamentally, it comes down to striking a balance between having a flexible deployment over a fast one, but with some work, you can have both.

There’s lots of ways this can be done. Applications can be installed by form factor. This could include rules your MDT ini that install VPN clients only on your laptops. Applications could also be installed by location, so that only desktops at a certain location get a certain program. This is also done by rules in the ini file based on gateway. Finally, you can also install applications based on the presence of specific hardware. Using some WMIC queries, conditions can be added to task sequences that install applications that are only needed when certain devices are plugged in.

That last example takes some work, you need to know the hardware IDs of the devices, and the devices would have to be standardized, but the point is that after some trial and error, You’re able to set up your imaging system to work for you, so that you can kick off a reimage, tell your user to go to lunch, and when they get back, they’re back to work, and you never get bothered again by them. Its truly a beautiful thing.

MDT can also be configured to detect and reinstall certain software during a refresh. During the gather phase, conditions can be set to create task sequence variables that would detect the install of software like Adobe Acrobat for instance. If MDT detects that Acrobat Pro is installed, the variable gets set and MDT will then install that application only on those PCs. The gotcha on this one is that the install has to usually needs to come from an MSI, but the framework exists, and could be used for any registry key if one wanted to go down that rabbit hole.

Build an image factory. Yes, this possibly one of the coolest things you can do with MDT. Some people don’t know this, but MDT is more than just an imaging system, its an entire automated systems deployment framework and as such it can be utilized to automate the entire image build process. This has allowed me to automate monthly builds of images, and keep our deployment times in the 25-30 min range.


Setting wallpapers, start menus, and desktop icons can all be scripted, and I’ve blogged extensively on some of these techniques, but the idea here is more of using MDT to build the configuration for you so that you know that the image you create tomorrow is more or less identical to the image you made last quarter just this one is fully patched, and that’s truly working smarter not harder.

Questions? Comments? Reach Out to Me Here!


Editing The Registry With MDT

I was helping somebody understand how to edit the registry with MDT, which isn’t very intuitive or obvious to some, and I ended up diagramming the entire commandline syntax of an example, and I figured it might help somebody else, so we’ll walk through it step by step on the blog today. First, you need to know what registry key you’re trying to set: In this example, we’re setting the lock screensaver option. This can easily be done via the command prompt like so:

reg.exe add "HKCU\Control Panel\Desktop" /v ScreenSaverIsSecure /t REG_DWORD /d 1 /f

What’s going on here? Look below.

Part of the beauty of task sequences is anything that can be done via commandline can be done via a task sequence. Simply Add it as a command line in your task sequence.

This is especially helpful when building an image. By using a combination of this and a few other tricks, I have fully automated by windows 7 image build so the hardest thing I have to do is run a task sequence which in turn builds and captures an image for me.


Ask MDTGuy: Adding Applications to MDT as .exe

Got an e-mail from a reader today about getting MDT to install a file as an exe.

In a perfect world, we’d all have MSI files to install software from because there is largely almost no guess work other than what paramaters to set if we need to. Sadly though, it is often the case with stand-alone exe files there is no documentation, and we’re playing the guess and check game.

Finding the magic switch to install software can be a real pain. The best thing to do is to try it on a PC first until you get it then add it to MDT. I like to copy the file to the desktop of a VM, snapshot it, and open a command prompt from within the folder. This is done easily by typing CMD into the address bar of the folder from explorer. This way you can nail it down first, then add it to MDT.

I usually start with the most obvious switches, /s, /q, -s or -q.

Sometimes however, you need to ask nicely and use /?, -?, /h or -h to get it to tell you what to do. This works more often than you think. It’s a way to ask the executable for help and in some cases it works, the package creator spent a few minutes leaving this behind for us sysadmins. For all the developers out there who do this, we salute you!

Microsoft installers are USUALLY pretty good about using /passive /norestart on stuff like dotNetFramework and other stuff, if all else fails, there is the googles. Usually putting in the name of the product or installer file name followed by silent or unattended will do the trick.

As Johan says, “Happy Deployments”!

What do you mean “PowerShell is required to use the Deployment Workbench”? Powershell IS installed!

Spent some time setting up a new laptop today and saw this error: PowerShell is required to use the Deployment Workbench. Found that to be weird considering I JUST used powershell to install MDT. Turns out I had forgot to set the execution policy back to RemoteSigned. Today, I learned: Powershell needs to be set to RemoteSigned the first time you run MDT.


Set-ExecutionPolicy RemoteSigned –Scope CurrentUser


Lenovos Not Identifying As Laptops, Chassis Type 31

Just got a pallet of fresh off the truck Lenovos in today, and of course if they don’t get the shiny new Windows 10 image here quick I won’t hear the end of it. Naturally, the first thing I want to do is image them, but I found it strange when I was prompted for the system type when assigning the hostname from the LTI wizard.

The wizard was showing %ComputerTypeName% and the serial with an error, that hostname was too long it said.

The only explanation was that for some reason these systems were failing to identify as laptops, as my customsettings.ini file is supposed to set hostnames based on a basic FormFactor and Serial Format. To make matters worse, this throws a wrench in all kinds of stuff like what OU and applications they get during deployment, so this was a problem that needed an immediate solution. Some googling led me to the wmic command to determine “chassistype” and I got a chassis type 31. Then I found this…

Chassis type 31 is “convertible” and I wondered, “Isn’t this all in the ZTIGather.wsf?” Sure enough, adding ,”31″ to the end of ZTIGather.wsf solves the issue….

It’s Friday, I’m going home. Its been one of those weeks, and next week is go time…

Update 4/6: Just got an e-mail from Paul at 1E mentioning I should “include a reference to table 17 in this document: It lists the official chassis type and thus gives us all confidence that we’re handling the new values correctly.” Thanks Paul!

Update 1/2018 99% sure this has been fixed in the latest build of MDT.

Decomissioning PCs with MDT

Now, I’ve known for a while there’s a way you can use MDT to wipe systems after you’re done with them. I know this isn’t some kind of KGB/CIA proof standard of wiping things, but it does the trick since I have PCs we were evaluating and I just need to send them back to the reseller. In this case, I’m dealing with systems that simply won’t boot to USB thanks to our wonderful secure boot technology, bargain bin USB flash drives so I need a task sequence based solution.

A while back I found this…

Googling the issue, I found this…

So, it really is pretty straight forward, just use the replace task sequence, and create a task sequence that’ll reboot into WinPE and wipe the disk. At first it didn’t work, buy

Create a Standard Client Replace TS
Note that by default there are conditions to running these steps. I added WipeDisk=True
Now, just browse to your share and run your shortcut to the wizard. Don’t run MDT from UNC? You’re doing it wrong.
Yeah, We got a task sequence for that!
WinPE is downloading! Reboot is Next!
Really I just wanted an excuse to show my WinPE wallpaper…

Okay, I know this wouldn’t fly at the DoD and all you tinfoil hat wearing mouth breathers are yelling “BUT THE GUBMENT CAN STILL GET MY DATA!” Okay, Okay, If three wipes of zeros don’t assure the Fox Mulder in the back of your head. We need to look into some serious data sanitation napalm. This is where sDelete comes in handy.

For you security types: check that out, it’s a SCCM task sequence, but it’s the same idea, no reason why it wouldn’t work with MDT. If you’re still that opposed to writing zeros or using sDelete and insist on using a DoD certified solution by hand, look into DaRT, Microsoft’s free¹ USB repair utility for Software Assurance Users.

¹ Free as in your employer pays lots and lots of money for a Volume Licensing Service Agreement or are a bad person that downloads software from bad places.

A connection to the deployment share (\\Servername\Share$) could not be made. DHCP lease was not obtained. Retry works just fine.

I hadn’t seen this one for a good minute, and I remember why. There are several reasons you can get this error. Sometimes you just need the driver. However the error screen will almost always tell you that you’re missing a driver, and no amount of clicking retry will fix that. Other times you’re genuinely seeing some kind of delay with the assignment of a DHCP address and fancy new hardware boots quicker than the network can get the systems IPs.

I know in the past I was able to add some kind of delay to get MDT to wait and then try to resume the task sequence, but I couldn’t remember what or where, so off to the googles I went…

I found this:

Then I remembered, yes, there is a delay we can add to the main heart and soul of MDT, the script of all scripts, the holiest of holies, the LiteTouch.wsf I hate doing this, I hate the idea of editing the scripts provided by Microsoft, because you never remember what you modified and where after an upgrade or after building a new share. But then again, you know what they say, “When in Rome Do as The Vandals!” Also, I’m tired of having to walk over to computers in our training room and clicking retry, so there’s that too.

Anyway, like Kyle describes in the abovementioned article, if it is a genuine delay in the DHCP assignment of an IP, we’ll need to add a delay in the LiteTouch.wsf script.


Look around line 1268. In this case, I’m adding a whole 10 seconds because I don’t want to come back and do this again if 5 seconds doesn’t cut it so I add wscript.sleep 10000 at line 1270. Lines 1269 and 1271 are optional snark that in theory will help me find it if I ever need to again.

Your Milage May Vary and as Johan says, “Happy Deployments!”