Best-Practices Remediation Tips for Server 2012 Pt I.

I'm calling this Part 1 because I realized as I started writing that this is a lot of work, and can easily be split into 2 or more articles.

Like most of the IT Pro community, I've been getting comfortable with Server 2012 the past several weeks now, and the journey is still ongoing. As I talked about last time, I do like those Best Practices Analyzers for Windows Server. Here's me running it in Server Manager:

Best Practices Analyzer Server Manager 2012*BPA in Server Manager 2012*

Getting any of these results back means that I have some work to do in remediating them. It's not uncommon for a Server 2012 system that was just built fresh with no applications loaded or configuration changes to still have one or two compliance issues in the Best Practices Analyzers. There is a balance to be maintained between compatibility and performance optimization. Also, many of these issues that popped up for me personally were not role-specific, but rather apply to a base component of the OS. Now I'll go over some of the interesting ones I've gotten and how I fixed them:

 

[Hyper-V]: Avoid storing Smart Paging files on a system disk.

Smart paging is new with Hyper-V 2012. Read about it here. Basically, you now enter a minimum amount of RAM a virtual machine can have, you enter a maximum amount of RAM a virtual machine can have, and you also now enter the amount of startup RAM a virtual machine can have. The Windows OS can boot up more comfortably with a larger amount of RAM, but once it reaches cruising altitude and is idle, the RAM requirements go back down, which will allow the Hyper-V host to gradually start reclaiming memory from the VM. With all this dynamic shrinking and growing of the memory on all your VMs, that's where the "smart paging file" comes in. And just like you can improve performance by putting your traditional Windows paging file on its own disk, the same goes with the Smart Paging file.

[Hyper-V]: Use RAM that provides error correction.

Microsoft doesn't support Hyper-V environments on hardware that isn't using ECC RAM. This is just a lab using desktop-grade hardware, so there's nothing I can really do about this. If you're using real server gear, this should not be an issue for you.

[Hyper-V]: Virtual machines should be backed up at least once every week.

This one I still don't understand. I have all the guest operating systems backing up nightly, and then I am backing them up again through the backup of the Hyper-V host. So go away, error.

[Windows]: Short file name creation should be disabled.

Then why is it enabled by default? Oh, I know why... it's because of your crappy line-of-business apps that were written back in 1998 that you can't get rid of, right? Well I'm in a pure 2012 environment right now, so I have no such worries. Good bye 8.3 filenames. Change this registry value to 1 to disable short file name creation on all volumes: HKLM\System\CurrentControlSet\Control\FileSystem\NtfsDisable8dot3NameCreation.

[Windows]: IrpStackSize should have the recommended value.

Irp stands for I/O Request Packet. Mark Russinovich did a great job of explaining IRPs and their role in the Windows I/O system in his book Windows Internals. All the various components or layers that an I/O packet traverses on its way to and from a disk, for example, are collectively referred to as "a stack." Each filter driver you add to the file system means that the IrpStackSize needs to be increased in order to accommodate it. A common example of this is when you install an antivirus product that uses a file system filter driver. If your IrpStackSize is set too small, certain operations might fail, such as attempting to access that machine's file system remotely. Conversely, it doesn't need to be set too high, either.  It was at 11 by default on my 2012 systems. The Best Practices Analyzer says it should be at 15, so I'll set it to 15. HKLM\System\CurrentControlSet\Services\LanmanServer\Parameters\IRPStackSize.

 

Alright, that's where I'll leave off for part 1. Part 2 will focus on some more network-specific optimizations, so check back soon!

Pingbacks and trackbacks (1)+

Comments are closed