Deploying Images with MDT/WDS

Posted on April 9, 2017 by jasonrw

The main point of MDT and WDS is to place Windows on a computer’s disk drive. To do this, MDT uses a series of steps in a task sequence that perform the necessary operations to facilitate installation. The crucial point to all of this is the network. Networking between the MDT/WDS server and the target clients must be clear and working. DHCP, DNS, and even WINS (if you’re feeling nostalgic) all must be working properly or else what could seemingly be an MDT/WDS problem, could actually be a network problem. MDT uses two ways to connect to the server over the network, USB key, and PXE. Pre-execution Environment (PXE) requires the use of a Windows Server configured with the Windows Deployment Services (WDS) role. MDT USB keys are copies of Windows PE, designed to connect to MDT and pull an image from the server. This is where the focus of this post will reside.

The good thing is that MDT has a provision for creating it’s boot sources built-in. Each time a deployment share is updated in MDT the drivers and settings contained within that share are added to a boot WIM and a boot ISO that is used to connect to the MDT server. These are located in the Boot folder of the deployment share. The WIM is used for copy to an WDS server and served-up via PXE. The ISO is meant for a CD-ROM, or the modern USB key. Simply booting a PC to that disc or key, and with DHCP working, the PC will connect to the MDT server. MDT does support the use of static IP addresses at the welcome screen, but DHCP is better and essential if PXE booting is desired.

Creating the MDT (Windows PE) boot volume…

Drivers are the biggest issue with regards to MDT’s Windows PE boot volume. Many times I have had to troubleshoot a connection to MDT that wasn’t connecting only to find out a driver was the issue. Like the target images, MDT’s PE needs network and system drivers to run on the clients to/for connecting. OEMs like Dell and Lenovo make Windows PE drivers available for their hardware, which makes it easy to organize those resources. Another place I go for network drivers is Intel’s web-site and their PRO LAN drivers.

Once I download and extract the drivers, I import them into MDT the normal way, but into their own folder within the driver store. Just like I create folders for the operating system and platform, I also create a Windows PE folder at the same level.

Once the drivers are added, I create a selection profile for just the PE drivers. This is so I can assign them to the MDT Windows PE boot device. Selection profiles are found under the “Advanced Configuration” folder, under task sequences. Simply, right-click the folder to create a new selection profile and drill-down the driver list to select only the Windows PE drivers. I create a separate profile for 32-bit and 64-bit platforms.

Now, from the properties of the deployment share, choose the Windows PE tab and then choose the “Drivers and Patches” sub-tab. From the drop-down list, any selection profile can be used with all or some of the drivers. This is where I choose my WinPE selection profile. With this, only the WinPE drivers will be used to connect to the MDT server. Note: there are two options on the WinPE tab, one for x86 and x64.

Other details can be set on the Windows PE tab, though the defaults are usually sufficient. For example, I change the wallpaper Windows PE uses to display a message on the screen, while the computer is imaging.

The image description for the WIM file is what WDS displays to the end user when they PXE boot to that server and are prompted to select a boot volume. I rename the WIM description and the ISO file name to follow suit with the deployment share.

When everything is as-desired, the deployment share must be updated to generate/regenerate the Windows PE WIM and ISO files. This process takes a few moments to complete, so grab a drink…

Once the deployment share has been successfully updated and has generated the Windows PE volumes in the desired configurations, we need to get those onto the WDS server, or portable media. Even with extreme customization, the ISO for the deployments share’s Windows PE volume is pretty small (by today’s standards). I’ve been using a legacy 512MB USB key for my 64-bit MDT Windows PE, which has to be over ten years old by now. Of course, later USB keys can be used. Use whatever method is best suited to place the ISO on the USB key as a bootable volume. I like to use a piece of freeware, called “Rufus“, but again use whatever works for you.

For WDS, the WIM file, created by updating the MDT deployment share, and placed in the “Boot” folder, along with the Windows PE ISOs, must be copied to the WDS server (if WDS, and MDT are on different computers), and used to “replace”, or create the WDS boot volume.

Go into WDS, right-click the desired boot image, and choose “Replace Image…”

After replacing the WDS boot image, it is not a bad idea to restart WDS on that server, just in case.

Deploy Time…

Get the target computer or VM ready by making sure it can either boot to the NIC, and get a DHCP address, or boot to USB. If you’re PC is using UEFI for its BIOS/firmware, booting to the NIC will most-likely have to be enabled in the BIOS setup (at least for Dell hardware anyway). Most of the PCs shipping from OEMs today will boot to a USB drive without any configuration needed. While in the BIOS setup, you might want to check that too. Pressing F12 on both Dell and Lenovo hardware will interrupt the boot sequence, and allow one to change the boot device for this particular boot, or enter the BIOS. Either way, boot to the USB to use a WinPE USB boot drive, or to the NIC for WDS. Also, keep in mind that complete network connectivity is required throughout the whole imaging process, from start to finish. If MDT’s WinPE volume cannot get an IP address, or see the computer/VM’s hard drive, the correct drivers need to be added. I think 85% of MDT imaging problems are driver-related.

I like to do my imaging off of the production network. First reason, is to not interfere with bandwidth that other computers may need during the day. However, MDT is a good network citizen. I’ve imaged dozens of computers over the production network, at the same time, without any negative impact. Second reason is that after imaging completes, you may have an unpatched copy of Windows on the network. This depends on the steps enabled in your task sequence, whether Windows Update is enabled or not, and how old the image being deployed actually is. Keeping the newly-imaged isolated from everyone else until it has had a chance to get the latest updates from WSUS, and get the latest virus definitions from SEP is a good idea. Since I am the only one using my MDT/WDS server, I keep that behind the NAT as well. My imaging traffic never even sees the production network, usually…

If MDT will deploy an image to a PC that already has a copy of Windows installed, and is not going to be performing an upgrade, wipe the target hard drive first. I use diskpart, and its “clean” command to quickly prepare disk drives on existing installations. Backup any and all data on the target PC before cleaning. If any doubt exists as to whether or not the data might be needed later on, back it up. Adding the following text to a simple txt file, called “clean.txt” will automatically clean any disk that is listed as disk zero.

select disk 0
clean

From a Windows PE command prompt, run the script by executing:

diskpart.exe /s clean.txt > diskpart_log.txt

Boot the computer, and connect to the MDT server. If things start going wrong at this point, there are few things, which are most-likely the culprit.

Can’t boot to USB drive, or the NIC? Check the BIOS, and modify the boot sequence, if necessary. Most new PCs are not set to boot to the NIC in the BIOS, so if that is your goal, check that setting and enable it. While in the BIOS, make sure ALL USB ports are enabled, and configured to allow boot. If the PC is set to successfully boot to either option, check the USB flash drive by using it to boot a known-good PC, which has been demonstrated to boot to USB or NIC in the past. For the hell of it, try a different USB port. I’ve only ever had one USB flash drive completely die on me. If the PC will not boot to the NIC, make sure that option is enabled in the BIOS, and that DHCP is working on your network. Make sure the WDS server is available over the network through ping via IP address and host name.

Successfully booted into MDT WinPE, but cannot connect to the deployment share? Start simple, make sure the WinPE session has received an IP address from DHCP. Press F8, in WinPE, and run the usual ipconfig command from the resulting command prompt. If DHCP is not available, find out why and fix it, or enter a static IP address at BDD Welcome screen (The first MDT screen, asking what is to be done). If a static IP address doesn’t work, then the problem is probably driver related. Add the drivers for the target hardware to the MDT driver store, and update the deployment share with complete regeneration (not the default option). Then, re-copy the WinPE ISO the USB flash drive, or replace the boot WIM on WDS with the updated copy, and restart WDS services on the server.

If the deployment share is accessed through a username and password, then make sure the correct set of credentials is being used. For domain use, I create a global group in active directory, and add those users that require access to the deployment share. Test access by having those users connect to the deployment share from Windows Explorer via:

\\server.domain.com\DeploymentShare$

If they cannot do that, they’ll never get in through Windows PE. Check the permissions… If the BDD Welcome screen, or a login prompt for the deployment share are not being displayed, the BootStrap.ini file is configured to bypass the welcome screen and automatically log users into the deployment share. That has to be done intentionally, and is not a default. Whether or not this is desired is up to the administrator of the deployment share. Having these steps pre-configured is essential for automating the imaging process, but not required for every MDT scenario.

Choose the option to deploy a new operating system on the BDD Welcome page, and select the desired task sequence. The steps given will all depend on how the deployment share’s CustomSettings.ini file is configured. I have one deployment share that is completed automated, without prompts. All the tech or whomever is imaging the computer sees is the actual process of MDT imaging the computer.

If the task sequence fails before it can re-partition and format the internal disk drive, there is most-likely a driver issue with either the chipset or storage controller in Windows PE. Again, add the latest WinPE drivers for your particular hardware. Even if you updated the drivers last week, check again. There may be an even newer update waiting. If that does happen, restart the image deployment from the beginning. Reboot the computer and “clean” the disk again with diskpart, or the new deployment will produce an error stating that a deployment is already in progress, and ask if you would like to continue that deployment, or start a new one. Chances are if the internal disk drive can be accessed, from Windows PE, by diskpart, and cleaned, there will be no issues accessing that drive to re-partition and format.

Once the task sequence has started, it will run by itself until something breaks, or asks for input. This again depends on the CustomSettings.ini, and specifically how the task sequence is configured. By default, all task sequences have an “Install applications” step, which will prompt the user to select which applications should be installed during deployment. Some task sequences will prompt for user data backup and restore, and some will prompt to capture an image of the computer, after deployment has finished. It all depends on how MDT is configured. If the deployment is running, just leave it alone and wait. Do not use either the computer receiving the image, or the one providing the image. For the wallpaper I use on my MDT Windows PE boot volumes, I have a Penn logo with a detailed inscription of what the computer is currently doing, and my contact information. The wallpaper image also asks that the computer not be used while the sign is present, and that interrupting the imaging process would require it to start all over again. Depending on network speed, the speed of the hardware at both ends, and the image size, the whole deployment process could be relatively quick, or could last a few hours. Just wait, and see until something breaks. If the task sequence fails in deploying Windows to that computer, MDT will indicate that with a pink-colored window. Completion, but with errors is indicated by a yellow-colored window, and success is indicated by a white window, indicating such. The MDT error messages are completely nondescript and far from helpful. The true key to identifying and fixing problems with MDT are the copious log files it generates.

MDT and Drivers

Posted on September 25, 2016 by jasonrw

Drivers are a big part of getting Windows to work properly. Anyone who has ever had to troubleshoot a piece of hardware on Windows knows this. When generalizing an installation of Windows for capture with sysprep, all but the most basic drivers are removed. This makes the captured image applicable to many different types of hardware. During a task sequence, MDT runs a plug and play check, for hardware at a couple of different points to determine what drivers, if any, are needed. When a matching driver is found, it is injected into the image before it is applied to the computer’s hard drive. To do this, MDT has a folder/section in the deployment share that is dedicated to organizing and storing drivers. It is called “Out-of-Box-Drivers” (OOBD).

The basic INF files are what MDT needs for driver injection. Many drivers are distributed as packages, which come in the form of an executable. This is not what we need. If an executable is the only way a driver is available, it must be imported as an application into MDT, and installed via task sequence. Fortunately, OEMs like Dell, Lenovo, HP, and even Microsoft make bulk downloads of model-specific drivers available from their sites. Dell hardware drivers come in the form of a CAB file, which can be opened with the expand command, or with a compression/decompression utility like 7-zip.

Bulk-driver Download Sites for Dell, Lenovo, HP, and Microsoft (Surface)

I create a folder structure on the MDT server similar to “Drivers\Windows ##\Vendor\Model.” For example: D:\Drivers\Windows 7\Dell\OptiPlex 990

I do not delineate between x86 and x64 versions of drivers in my folder path because nearly all of my OS deployments are 64-bit. Dell combines x86 and x64 drivers in the same download. Many Dell drivers can be used on both platforms. Lenovo drivers will try to extract to their own, specified path, but that can be changed at runtime. Complete, downloaded driver packs can be between 300MB and 1,000MB in size, except for WinPE driver packs which are very small.

Using the expand command, I’ll extract the drivers to my folder structure.

expand C:\Users\jasonrw\Downloads\990-OptiPlex-ABCDE.cab -f:* “D:\Drivers\Windows 7\Dell\OptiPlex 990”

The command prompt window will scroll really quickly and end with the prompt returning. Now, we can import them into MDT. MDT is able to handle drivers in different manners. The basic default option is to throw all drivers into the same folder at the root OOBD directory. With that, a deployment task will search the entire OOBD store for the right drivers. This increases the chance of the wrong driver being selected. WMI is great, but it is not perfect. Another option is to break down the OOBD store by manufacturer/vendor or by operating system version. Creating folders for Windows 7 and Windows 10, respectively can help minimize the chance of a wrong driver being installed. I take it one step further, actually a couple of steps further. I still use MDT’s WMI hardware-querying capabilities, but I tell it exactly where to look.

I create a specific folder structure under OOBD that matches a specific WMI query I pass on to the deployment task. For example, Windows 7: Out-of-box-drivers\Windows 7\Dell Inc.\OptiPlex 990

The bottom two folder tiers each correspond to variable in a WMI query, Make and Model. To find the make or manufacturer of an OEM PC, run the following command from a command prompt.

wmic computersystem get manufacturer

The returns for Dell and Lenovo are “Dell Inc.” and “Lenovo”

To get the model, run the same command as above, but replace “manufacturer” with “model” Top that off with a folder for OS version and platform, and you have something to use. In the task sequence, a task variable can be inserted into the PreInstall phase, before the inject drivers step to tell the task sequence exactly where to look. The variable is “DriverGroup001”, and the value is “Windows 7 x64\%Make%\%Model%” This will allow a task sequence to correctly use a WMI query to find the drivers for the exact make and model being imaged.

The Inject Drivers step has its own settings that needs to be configured. For our purpose, the selection profile has to be set to “Nothing” with all drivers from the selection profile being used.

Selection profiles are the ultimate step toward driver control. They are a pre-defined selection of drivers that may encompass and individual model, or manufacturer. MDT ships with a few pre-defined selection profiles, but more can be created to suit any need. Given the exact control this approach provides, there is one detraction, it limits task versatility. Since a selection profile tells a task sequence exactly which drivers to use, MDT doesn’t query for them. The default setting for Inject Drivers is to query the entire OOBD store, but when “Nothing” is set, they querying is off. If it is desired that a task sequence only serve one or two makes of computers, this might be a good approach. I support about a dozen different models from two manufacturers, and I’d like my task sequences to be applicable to all.

I do use a selection profile to organize the drivers I use for my MDT Windows PE ISO/WIM files. Dell and Lenovo make drivers just for Windows PE available as a download too. I use the same approach as above to download, extract, and import the Windows PE drivers into MDT. Then, I create a selection profile for the Windows PE drivers and use that for my ISO/WIM file drivers.

To import drivers into OOBD, make your folders as desired, right-click the folder for the computer model, and choose “Import drivers.” A wizard will open, walking you through the process of importing the drivers from where they were extracted. It is real easy.

After the driver import, the deployment share must be updated. By default, MDT uses the all network and system drivers it has in the OOBD store for the Windows PE ISO/WIM file. This can be changed, as I described above, with a custom selection profile, but it is not mandatory. Still, note, that each time drivers are added and removed, the deployment share should be updated for those new drivers to be used. Some drivers are clearly depicted whether or not they are x86 or x64. In reality, many single drivers can be used on both platforms, but the descriptor files do not always indicate that to MDT. Thus, MDT will import the driver and override the specified platform. This is noted after all of the drivers have been imported for that operation.

I, personally, write down the name of each driver with a warning, and disable them after the wizard ends. Disabling/deleting a driver is easy, just right-click it in the MDT workbench and choose the appropriate option. Drivers must be disabled from their property sheet. Disabling a driver is a safe approach before it is determined that deletion is necessary. Only ever delete a driver from the MDT workbench. DO NOT go into the driver store via the file system and delete it that way.

Again, when disabling or deleting a driver, you must update the deployment share to take it out of Windows PE.

Finding, downloading, extracting, and importing drivers into MDT is a big part of MDT configuration, which takes a great deal of time. If it is done with forethought and planning, it can minimize the driver problems a deployment share might have, and need only be done once. I note the name and date of the driver files that I download and import into MDT. Then, I can periodically check for updates from the vendor’s web-sites. The older a model is with the manufacturer, the less-frequent they tend to update the drivers packs for that model. If the manufacturer does not make a driver pack available for your model, it is possible, though very tedious, to download each driver, and extract them individually. I try to avoid doing that.

Thanks!

UEFI and a Dell OptiPlex 990

Posted on April 7, 2016 by jasonrw

Hello,

I do a great deal of work with virtual machines and perform all of my operating system development on virtual platforms. My desktop PC came with a 500GB hard drive. Using a virtualization program and creating a couple of production sized virtual machines will take up a great deal of that space very quickly.

A larger hard drive is one of the easiest upgrades one can make to a computer. I ordered a 3TB drive for my Dell OptiPlex 990 and had hoped to just plug and chug, but it didn’t work out that way. What I had found out that the default traditional method of BIOS hardware management on the PC only allowed it to see partitions or drives no larger than 2TB. BIOS-based computers use hard drives that are partitioned in the Master Boot Record or “MBR” format. To use the new crop of large drives with more than 2TB of space, one needs to format the drive as a GUID Partition Table or GPT device. There, the largest of drives can be partitioned and formatted as one single volume, which is what I was after.

I should have just done my homework and just ordered a 2TB drive. Under BIOS/MBR, Windows setup would only see 2TB, leaving around 768GB unavailable. GPT disks cannot use the BIOS system. They rely on a newer system called the Universal Extensible Firmware Interface, or “UEFI.” The OptiPlex 990 has the ability to use a BIOS-based system or UEFI, but not both. Each are mutually exclusive to the OptiPlex 990. Newer computers have the ability to use UEFI with support for legacy BIOS system (UEFI-CSM), but not the 990.

I initially tried to get the 3TB HDD to be recognized by booting to a Windows PE boot drive and use DISKPART to partition and format the HDD as a GPT disk. Still, Windows setup would not use the drive. Windows 10 setup would indicate that setup could not use any available partition as they were in GPT format. The problem wasn’t the drive, which was set up correctly, it was the install media. I use an 8GB USB key drive to install Windows from an ISO file. A great piece of freeware called “Rufus” is what I use to make the boot-able USB key from the Windows ISO file. Rufus defaults to creating a Windows PE volume, which is what Windows setup is, that supports both MBR and UEFI with backwards compatibility. I though that would work, and it should have, but the OptiPlex 990 was BIOS or UEFI, not both. One option to create the USB boot key was pure UEFI.

While I was sorting this out, I noticed that when I switched the 990’s BIOS to UEFI mode, there were no more boot-able drives like the HDD, CD/DVD, or USB listed. This was accurate because none had been made available. The 3TB drive was GPT but did not contain any boot-able volumes at that point. I had yet to make a pure-UEFI USB key, and there was no disc in the DVD-RW drive, so no there were no UEFI boot options. Any and all BIOS/MBR, and UEFI/GPT boot devices are shown when the PC POSTs. As soon as I partitioned and formatted the 3TB drive as GPT, and inserted the all-UEFI USB boot drive, the OptiPlex 990 saw the USB drive as a UEFI boot device and Windows setup accepted the partitioned 3TB drive for install. After that, Windows 10 installed and I had ALL of the available space, which came in around 2.78TB.

To do…

Partition/format the drive as a GPT device.

Create a boot-able Windows PE USB drive and boot the target computer to it, with the larger-than-2TB-HDD installed.

Run DISKPART from the Windows PE prompt and enter the following commands to partition and format the drive as a GPT disk, minus the REM statements.

select disk 0
clean
convert gpt
rem == 1. Windows RE tools partition ===============
create partition primary size=300
format quick fs=ntfs label="Windows RE tools"
assign letter="T"
set id="de94bba4-06d1-4d40-a16a-bfd50179d6ac"
gpt attributes=0x8000000000000001
rem == 2. System partition =========================
create partition efi size=100
rem    ** NOTE: For Advanced Format 4Kn drives,
rem               change this value to size = 260 ** 
format quick fs=fat32 label="System"
assign letter="S"
rem == 3. Microsoft Reserved (MSR) partition =======
create partition msr size=128
rem == 4. Windows partition ========================
rem ==    a. Create the Windows partition ==========
create partition primary 
rem ==    b. Create space for the recovery image ===
shrink minimum=15000
rem       ** NOTE: Update this size to match the size
rem                of the recovery image           **
rem ==    c. Prepare the Windows partition ========= 
format quick fs=ntfs label="Windows"
assign letter="W"
rem === 5. Recovery image partition ================
create partition primary
format quick fs=ntfs label="Recovery image"
assign letter="R"
set id="de94bba4-06d1-4d40-a16a-bfd50179d6ac"
gpt attributes=0x8000000000000001
list volume
exit

Script from Microsoft

Create the all-UEFI Windows install USB drive…

Go into the computer’s BIOS by pressing F2 during the boot sequence (Dell), before Windows even starts to load. NOTE: BIOS/MBR disks will not boot in a UEFI/GPT configuration. Any change from BIOS/MBR to UEFI/GPT WILL REQUIRE a Windows reinstall. No way around it, so Back up your data.

On a separate computer, insert the USB key, and open Rufus as an admin.

The USB key should be listed in the top dialog box, make sure you’re formatting the right drive if there are multiple USB drives currently inserted into the PC. From the bottom part, near where it says “Create boot-able disk using [ ISO Image]”, click the small button with the disc and drive icon and choose your Windows install ISO.

Next, select the drop-down menu option under “Partition scheme and target system type”, choose “GPT partition scheme for UEFI.” If the source ISO file changes, the partition scheme changes also, so watch that.

Create the drive and insert it into the computer that has the large HDD and boot to it by pressing F12 (Dell/Lenovo). If the USB key was created properly as a UEFI device, it will show up as a UEFI boot option under the BIOS boot options which will still be the CD/DVD drive and the HDD (possibly).

Install Windows and relish in the large space now available!

How to Create a Windows Image for Mass Deployment

Posted on March 1, 2016 by jasonrw

Requirements: Windows install media (7 or 10. 8.x?), desired apps for the image (Office, PDF viewer, web browsers, plugins), virtual machine software (VMware Workstation, Microsoft Hyper-V, or Oracle Virtual Box), and image creation and deployment software (ImageX.exe, MDT, SCCM).

Almost every place I have ever worked, IT had or needed a method to clone and deploy a specific Windows configuration and application set. From a few PCs, to hundreds, the requirements were the same, to deploy the same configuration with as little, repetitive work as possible. The ideal target being what Microsoft calls “zero-touch” deployments that require no interaction on the target computer whatsoever. This is offered by Microsoft System Center (SCCM) along with the Deployment Toolkit (MDT). Many shops do not operate that way, and have some level of interaction required during the imaging process. This piece will discuss creating a Windows install for distribution.

What you’ll need…

Windows and software install media (obviously)

Virtual machine software for the creation workspace. Why? Two reasons. First, virtual machines provide the option to create hardware-neutral images which can be applied anywhere, regardless of what is actually in the target computer. One image becomes possible for multiple hardware configurations. This also involves less work in mainatining the image as any work only needs to be done once and not x-times per different type of hardware. Second, most virtual machine software (I’m not sure about Virtual Box) have the ability to save a VM’s state, and revert back to that state, should it become necessary. VMware calls these “snapshots”, and Microsoft uses the term “checkpoint” in Hyper-V. Should a screw-up occur, it can be undone without loosing work or have to re-do everything. These are two facets that are simply not available with building images on real hardware. Test on real hardware, but build in a virtual environment.

VMware Workstation is pricey, but well worth the cost IMHO.
Hyper-V comes with Windows Server 2008 and later, Pro and Enterprise versions of Windows 8, 8.1, and 10 as “Client Hyper-V.”. The build computer’s CPU must support hardware assisted virtualization for Windows to install the Hyper-V role. Intel Core 2 Duo/Quad CPUs won’t muster.
Virtual Box, now owned by Oracle, is a freebie. I haven’t used Virtual Box very much outside of general curiosity.

The build workstation has to have some power to it. Nothing extravagant like an Alienware, or Falcon Northwest gaming rig, but above average. Try to avoid using a laptop as a VM build station. Laptops are great for testing, but a desktop PC is optimal. Don’t use a Mac. I love my MacBook Pro, but it isn’t meant for making Windows images. A quad-core CPU (Intel Core i5/i7, or AMD Phenom series) will work for starters. The more powerful, the better. RAM is the key. The more the better. I routinely work with 16GB of RAM on my workstation (the most it’ll take), and it can handle three running VMs and the host OS before going wacky. 32GB of RAM is not ridiculously expensive today, and well-worth the couple-hundred extra bucks. VMs take up storage space quickly. Working on several VMs, it is not difficult to fill a 2TB HDD (I’ve done it). Those are not that expensive either, and 2TB is the starting point I’d go with for a virtualization rig. Anything more, and you have to make sure your PC supports UEFI vs. BIOS, or else all of the drive’s space will not be recognized by the firmware, and Windows. Working from USB storage might fly, but the throughput won’t match that of internal storage, and you’ll have a bottleneck. My VM creation setup, however, is backed-up every night to my trusty 4TB WD USB HDD. If you can get large-capacity SSDs instead of traditional rotational drives, do it, but don’t sacrifice space for speed. An SSD for the boot volume with Windows and apps along with a large 2TB+ traditional HDD for VM storage is a nice setup, and not fiscally unrealistic.

Virtual Machine Setup…

Create a new VM that will become your Windows image. For Windows 7 and later, I recommend 4GB of RAM, 1 CPU with 2 virtual cores (if possible), and a virtual hard drive the size of the smallest drive that will ever receive the image. We have some 128GB SSDs out there, so 128GB is my vHDD size. That’ll assure the image will fit everywhere you intend to deploy it. Now, it is easy to see how fast space will go on your drives. Everything else is fine with the defaults. Bridged vs. NAT network adapter? It doesn’t really matter even for network-based capture/deployment. I’ve used both and have noticed no speed differential.

Install and Configure Windows…

Install Windows onto the VM with all of the default settings. If you’re lazy like me, you can use an unattend file to answer all of those pesky setup questions. This site, the Windows Answer File Generator, has a GREAT web UI for creating unattend.xml files that WORK. Place the unattend file at the root of your install media and let Windows install itself. The whole process for Windows 7-10 should be 20-30 minutes. Don’t bother with the product key and activation. Sysprep will just strip that out in the end. If your build process takes longer than a month, you might need the key and activation. I’ve never run into that problem. Don’t join any active directory domains. Sysprep will quit if it is run on a domain-joined PC.

Next, power-down (not sleep, hibernate, or pause) the VM and create a snapshot or checkpoint. This will save you 20-30 minutes of re-installing should a foul-up occur. Power back on and go into Programs and Features (Windows 7) and add/remove all of the stuff that is not needed. I get rid of the tablet components, XPS printer/viewer, Windows Media Center, Windows Fax and Scan. All of the stuff I know the end users will not touch.

Update Windows/Office…

Before getting into the nitty-gritty of configuration, completely update Windows through Microsoft Update. The older your version of Windows, the more updates it will need, and the longer the update process will take. A fresh install of Windows 7 Enterprise SP1 x64 from MS VLC ISO required 296 updates before no more were required (as of February 2016). This will take about a day (8+ hours) to complete. One might think about adding MS Office into this process to allow it to join in the update process, but there is a reason not. Take this opportunity of having a clean install of Windows, updated, and use it as a template for other VMs. VMware Workstation allows VMs to be cloned and copied, so a patched copy of Windows can serve as a starting point for other virtual machine projects. A MAJOR time saver. Completely update Windows until it screams “no more” and shut down, then take a snapshot.

Clone or continue? That’s up to you, but whatever is chosen, the next part is installing MS Office (if needed. I can’t imagine it wouldn’t), and completely updating that through Microsoft Update. I do a custom install of Office to not include the programs users won’t need, like Infopath, and Lync, Skype for Business, and OneDrive for Business. YMMV. Microsoft Office 2013 Pro Plus needed 151 updates to be complete from a fresh install of the MS VLC ISO. Again, I wouldn’t bother with product keys or activation for Office. Shut down and take a snapshot when that part is done.

Application Installs…

Add all of the applications that need to be deployed with the image. Here is where the question of thin image vs. thick image is contemplated. A thin image contains just the bare essentials needed to get started with many other apps installed at distribution. Programs that change often, and would require updating/re-capturing the image are best left to installation at deployment. Candidates for this include Adobe Flash Player, Shockwave, AIR, and antivirus software. Software that has first-run settings which must be answered for the end user should be placed into the image, and not installed at deploy time. We install many apps with our image, so this process takes us about as long as it does to update Windows and Office. A good example set to start would be: Mozilla Firefox ESR, Google Chrome Enterprise, Adobe Acrobat Reader, Quicktime Player, VLC Media Player, Microsoft Silverlight, Visual C++ Redistributables 2005-2015, Skype and 7-zip. Run each and every newly-installed application to make sure they work as intended, and then delete the downloaded installers. The plugins, and antivirus (SEP) will be installed when the image is deployed. Power-down the VM and take another snapshot.

By this point, we have a basic working install of Windows which is moderately useful and could probably be distributed to end users. One question does arise which is substantial in nature, and determines how next to proceed. Does a custom default user profile need to be created and configured? If yes, we need to do that before capture.

Customizing a default user profile…

Windows, by itself, works pretty well out of the box, but comes with a myriad of first-run dialogs and prompts which can be confusing. My target audience is public computing, classrooms, kiosks and labs. For privacy reasons, user profiles are not kept on the computers. As soon as a user logs off from Windows, their profile is removed. Each time someone logs into one of the computers I am responsible for administering, they are logging in for the first time. Any and all first-run dialogs/prompts that can appear will. The option to spend 10-15 minutes of a 60 minute class, getting the software to work as desired, and out of the way is just not an option. To prevent this, I try to configure as much for the end user in advance as possible. Group policy is my hero in this effort. Almost anything can be set for the computer or a user through a GPO. The exceptions being non-Microsoft software like MATLAB, Maple, and Stata that all have first-run issues which often require administrative intervention. I don’t let users run as admins.

Sysprep is the utility Microsoft has made available for generalizing an installation of Windows since Windows 2000. Starting with Windows Vista, Microsoft changed to an image-based installation and mmaintenance process. Sysprep changed too, and became much more difficult to use (for me anyway). It is at this time that copying a customized user profile to the default required the use of an unattend.xml file. With Windows 2000 and XP, you could actually just copy it and everything would be fine. In the specialize pass (No. 4) of the unattend.xml file there is an option to add a pass called “CopyProfile” to the “Microsoft-Windows-Shell-Setup” setting that will copy the built-in administrator account’s profile to the default. The trick is when to apply this, during capture, or during deployment? That depends on how you’re capturing and deploying, but either way the built-in administrator account is what I use as a template for my custom default user profile.

Enable the built-in administrator account and give a password you’ll be comfortable with entering dozens of times. Log in as the admin, and immediately delete the profile for the other account setup asked you to make after installing Windows. The reason for this is that CopyProfile will not copy the admin’s profile if another profile exists on the file system at the time the copy takes place (experience speaking here). Delete the profile from the advanced system properties window, and not by just deleting the folder under C:\Users. Other accounts can remain, they just cannot have profiles.

As the built-in admin, configure the Windows environment they way you want the end users to have it. Again, most of these tasks can be accomplished with a domain-based GPO. Run each and every program the end users are likely to use and answer any first-run dialogs and prompts. DO NOT surf the web as the built-in administrator on a Windows PC without antivirus software for obvious reasons, but an important another is profile size. Surfing with Chrome or Firefox even on just a couple of sites and will add megabytes of data to the profile. You could probably get away with not even running any of the browsers at all with a good GPO in place. Google, Mozilla, and of course Microsoft make GPO settings available to completely configure each piece of software and eliminate any first-runs. Adobe recently made an admx template for configuring Acrobat Pro and Reader via GPO. GPO templates exist for Microsoft Office 2007-2016, with a dizzying array of possible configuration options. Those and the settings for Windows will get 95% of end user configuration done in a customized profile.

Things that are easily set include the desktop wallpaper and icons, the start menu, screensaver, power options, and remote desktop settings. Take your time and run through each usage scenario, if possible, without puffing up the profile’s size. Large user profiles take a while to create at log in, and lengthen the log in time required to get started. In my case, every is logging in for the first time, so their profile will be created when the need to use the PC. A 500MB profile will slow that entire process down. Once things are as desired, power the VM down and take a snapshot. This snapshot is a failsafe point for return after the image has been captured.

Capturing the Image…

There are several different ways to get an image of Windows. Traditionally, Norton (Symantec) Ghost was the standard for deploying Windows operating system images. After acquisition, Symantec let the product stagnate over a period of years as Microsoft developed successive versions Windows, and it became necessary for us to switch to a solution that would natively support later versions of Windows PE. We adopted MDT a couple of years ago for a few reasons, and that allowed us to change the way we made operating system images. Driver support in Ghost is not real versatile. The option to create one image for multiple hardware configurations required substantial tweaks and endless trial and error testing Ghost was designed to have one image per hardware type, with all of the drivers included in the image. This is not that big of a deal since many large IT outfits only support a few different types of hardware models and configurations. For us, that was six different images for a single type of Windows install (Windows 7 Pro x64). Needless to say the images did not get updated too often due to the amount of work involved.

Removing the hardware dependence and creating hardware-neutral images was a requirement for our new imaging software. An install in a virtual machine allows that type of neutrality.

Boot the VM to be captured an log on as the built-in administrator, the one that was pre-configured before. From the advanced system properties, delete every other user profile on Windows. The accounts can stay, but the profiles cannot. If there are any VM-dependent software like VMware Tools, or Hyper-V Integration services installed, uninstall them. If there are mapped drives between the host and guest OS, power-down the VM, remove them, and restart.

I like to clean Windows before capture by running a few command prompt executables.

Go to %TEMP% from the Run dialog and delete everything there that can be deleted.

Open an administrative command prompt and run the following commands.

Delete any and all shadow copies.

vssadmin delete shadows /All /Quiet

Get rid of any downloaded software updates.

del c:\Windows\SoftwareDistribution\Download\*.* /f /s /q

Delete any hidden Windows install files. Chances are there are none, but it cannot hurt to check.

del %windir%\$NT* /f /s /q /a:h

Delete the Windows prefetch files. There also probably none of those either.

del c:\Windows\Prefetch\*.* /f /s /q

Run disk cleanup.

c:\windows\system32\cleanmgr /sagerun:1

Defragment the C:\ drive (It shouldn’t be that fragmented).

defrag c: /U /V

Clear the Event Logs. Execute one command on each line.

wevtutil el 1>cleaneventlog.txt
for /f %%x in (cleaneventlog.txt) do wevtutil cl %%x
del cleaneventlog.txt

Flush the DNS cache.

ipconfig /flushdns

NOW! We’re ready to capture. The question is how?

We use MDT for imaging. MDT has a special type of task sequence called “Sysprep and Capture.” To kick this off, from the install of Windows to be captured, navigate to \\mdtserver.domain.com\DeploymentShare$\Scripts and run LightTouch.vbs. This will connect to the deployment share, and start the process. Enter any credentials required, select the appropriate task sequence, and give the image a name, then begin. Capturing from a VM to an actual MDT server, over the network, will take a while. Even for small images, it is best to just let the task run and not use the build computer for anything else until it is finished. I do this at the end of the day, when I’m not going to be needing to use the computer.

Once the image has been captured, the VM will restart and wait for further action. At that point, power-down the VM, and roll back to the last snapshot taken before the VM was captured. Back to not-so-square-one, and ready to re-capture, update, or whatever when necessary.

Outside of MDT, it is possible to capture with just a Windows PE boot disk with ImageX.exe. This process is not a clean and automatic as MDT, but it works. Going that way, the pre-capture setup process goes a little differently. In its most basic form, you need to only run sysprep from the VM about to be captured, and shut down, then restart to the WinPE boot disk and run ImageX.exe. To copy the default user profile this way, an unattend.xml file needs to be used with sysprep and the CopyProfile option must be set to “True.” The unattend file is only needed if there are any customizations that need to be applied when the image is applied to a computer. Large capacity USB flash drives are very affordable. I have a 128GB USB drive, that I purchased for $29.99, configured as a WinPE boot drive, and can be used to capture images directly to the drive instead of over the network, as it is usually done.

Enjoy!

MDT, Getting Started

Posted on October 30, 2015 by jasonrw

The Microsoft Deployment Toolkit (MDT), a very nice option for deploying Windows in small to large environments.

Symantec Ghost was the de facto tool we used to deploy Windows to end user computers. One big issue was that Ghost, as a product, wasn’t updated to keep pace with the new versions of Windows. Another issue with Ghost was driver management. There was no easy and reliable way to configure a central driver store that the images could use, so multiples images, one for each hardware type, had to be kept. MDT uses a shared, central store to keep drivers. Initially, Ghost was able to automatically reboot client into an image task via a desktop agent. As Windows progressed in development and Ghost stagnated, the compatibility with Windows PE, the tool used by Ghost to start a computer into an imaging task, became nil. To use Ghost, at the end, each client computer had to be manually started into an imaging task with a USB boot key.

The requirements for any new imaging system included a hardware-neutral image creation process, a central driver store, and the ability to remotely initiate an imaging task on a client computer (zero touch). Two out of three isn’t too bad. MDT offers the ability to use a hardware-neutral imaging process with a central driver store, but it doesn’t specifically offer zero touch imaging. The higher-end product, Microsoft System Center, does. MDT is centered around a lite touch approach which requires a minimal, and manual interaction with the target desktop. Many settings, however, can be automated to the point of complete autonomy. Best of all, unlike Microsoft System Center, MDT is free of charge.

To use MDT version 2013 Update 1, Windows 7 or better is needed as a source platform. I tested with a proper version of Windows Server 2012. Windows Server is required at some point for PXE-booting capability via Windows Deployment Services (WDS). WDS will allow a PC to boot to a network-based Windows PE image, created with MDT, and commence imaging. Windows Server will also support more than ten simultaneous SMB connections unlike a desktop version of Windows. An essential part of setting up MDT on a server is the Windows Assessment and Deployment Kit (ADK) for Windows 10. The ADK contains the tools necessary to create the deployment infrastructure for MDT, most notably Windows PE.

For example:

Install Windows Server 2008 R2 or better and update it to it’s fullest extent (that may take a while), in addition to installing antivirus software and joining an active directory domain (optional).
Then, install the ADK for Windows 10, which covers deployment for Windows 7, 8, 8.1, and 10 as well as the servers. If there are no plans to use the MDT database with the server, SQL Server Express need not be installed with ADK. SQL can always be added later if the requirements change. Reboot the server.
Install MDT 2013 Update 1 on the server. If you’re unsure over which version of ADK or MDT to install, go with the x64 version. In my experience, the x64 version will deploy a 32-bit (x86) operating system, if need be.

ADK for Windows 10 installing on Windows Server 2012 R2

Now, MDT is set up on a server but it cannot do anything yet. There is much more configuration necessary to get to a working imaging and deployment solution. The first part is to create a deployment share. Everything MDT does is from within one or more deployment shares. Since MDT is very similar to a dedicated file server which shares Windows images and applications, it is best to use a very large storage space for the deployment share. I tried to set up a deployment share on a separate DFS path from the server’s own file system, but it did not work. Windows PE couldn’t find that deployment share during a deployment task sequence. For testing, a 1TB volume was more that sufficient for test and low-level production use. I’d split the 1TB volume into two partitions, one for the operating system, and the other for MDT’s deployment shares.

Should I dedicate an entire server or VM to MDT? It really depends on how much use is planned for it. Even if MDT is used to create and pull custom images of Windows which contain only the bare minimums necessary to create a usable, working Windows install, those images will start to take up space. Use as a traditional file server in addition to MDT is practical, but there may be performance deficits during times of peak imaging. YMMV…

Another important thing to note is that at this point, before the MDT config is set up, make sure the host name for the server is set to the desired name. Windows PE uses the server’s host name in the UNC path to connect to the deployment share during an imaging task. If the server is renamed, post MDT-config, those changes will NOT propagate to the deployment shares. It is not that much of a big deal to modify the deployment share to use the new name, but it is a hassle and best avoided.

To create a deployment share, open up the MDT Deployment Workbench as an administrative user on the server. It might even be a great idea to pin a shortcut to the Deployment Workbench to the task bar or desktop. The deployment workbench is divided into two sections, the Information Center, and the Deployment Shares. Choose the lower-half of the window, and right-click Deployment Shares and choose “New Deployment Share.” A wizard will appear, walking you through the creation process.

The first part of the wizard is to establish the local file path for the new deployment share. The default is “C:\DeploymentShare.” I would make this on a volume that is not shared with the operating system and that has plenty of space available. The new share can be named anything, but it cannot and should not ever be renamed. What name is chosen here is what will be for that share for it’s duration. Since I’m just testing here, I’ll go with the default.

The next step establishes the new deployment share’s share name, which is just the chosen name of the folder with a dollar sign ($) appended. In this example, the share name is “DeploymentShare$.” This share must be freely available on the network to clients that will be imaged with MDT. As stated before, the name of this share should never be changed once it is established.

After that, the next step is to give the new share a descriptive name. Again, this can be anything. If multiple deployment shares will exist on the same server, this name will be used to differentiate them in the Deployment Workbench. Be succinct and descriptive.

The Options page prompts for information that will form the basis of the deployment share’s configuration file, CustomSettings.ini. Any setting chosen here can be changed later on by editing that file. Usually, the defaults are fine for starters.

The next screen is a summary of the options chosen up to this point. The opportunity to go back and change anything is available here. Once this the wizard has advanced past this screen, the option to go back is no longer available. Make sure everything is as desired and proceed. The wizard will create the share and upon a successful completion will present it in the Deployment Workbench.

The new deployment share in the Deployment Workbench

Despite being created, the new deployment share cannot do anything yet. In order for it to become functional, the new deployment share needs one or more operating system images, applications to deploy at the time of imaging (optional), drivers for the deployments, and task sequences to coordinate the deployment process. The deployment share is divided into several folders, each named after it’s intended function. A right-click to any of those folders will provide the option to add content to those folders. Sub-folders can be added under the main folders for further organization. It is a good idea to divide your operating systems folder by the OS being used, the drivers folder by the manufacturer and model of computer holding the drivers. It doesn’t have to be done that way, but organization provides further opportunities for customization down the road.

The deployment share is controlled by a configuration file, a text file called “CustomSettings.ini.” The file resides in the deployment share’s folder, in the Control folder. CustomSettings.ini (CS) can be accessed easily by right-clicking the deployment share in the Deployment Workbench and choosing “Properties” and selecting the “Rules” tab.

A default CustomSettings.ini in the deployment share properties.

A default CustomSetting.ini in the deployment share properties.

Bootstrap.ini is a configuration file that is placed on the Windows PE volume when the deployment share is updated. Some of the configuration changes made to a deployment share are incorporated into Windows PE once the deployment share is updated. Upon update, the Windows PE volumes are created with the Bootstrap.ini file, drivers and any other selected options. The deployment share need only be updated when drivers are added/removed from the deployment share, and changes made to the Bootstrap.ini file. To update the share, right-click it’s folder in the Deployment Workbench and choose “Update Deployment Share.” The first update will take a while to complete as all of the drivers, or just the selection preset, are added to Windows PE. After that, the Windows PE ISO is only updated with the most recent changes made to the deployment share, unless the share (Windows PE) has been selected to be completely rebuilt.

Jason Watkins

10/30/2015

The next step is located here.

Note: Microsoft has a great series of Technet articles on how to deploy Windows 10 with MDT. The information contained in this post is a much simpler version of that information.