One of the things I’ve maintained over the years is a lab to do everything I need. From testing new software, proof of concept designs, benchmarking or testing new builds of SQL Server.
This has always been a tall order and now that I’m back developing software part-time I need a solid, yet cheap, lab to do mostly functional resting in.
To give you an idea of how tall an order testing software to run on SQL Server has come let’s take a small trip down memory lane.
Once upon a time…… I was the product manager for Litespeed at Imceda, life was simpler then. We had a small support matrix.
SQL Server 7
SQL Server 2000 32bit Itanium
Windows NT 4
Windows 2000 Itanium
Windows 2003 32bit, Itanium, 64
Clustering on all platforms except Itanium
Multiply that by the growing feature set and it added up to hundreds of tests that needed to be executed for any built that would see the light of day.So, we setup several machines some virtual to handle the functional testing, some physical to do performance testing. We had automated test harnesses to do some of it but there was still quite a bit of manual work to be done.
Fast forward to today and this is what my test matrix looks like:
SQL Server 2000 32bit Itanium
SQL Server 2005 32bit 64bit Itanium
SQL Server 2008 32bit 64bit Itanium
SQL Server 2008 R2 32bit 64bit Itanium
Windows 2000 32bit Itanium
Windows 2003 32bit 64bit Itanium
Windows 2003 R2 32bit 64bit Itanium
Windows 2008 32bit 64bit Itanium
Windows 2008 R2 32bit 64bit Itanium
Clustering on all platforms except Itanium
Yikes! This is getting huge. This is starting to feel like testing for Oracle and Unix platforms with the only exception you can’t pick and choose what to test. If it is under support by Microsoft you are pretty much stuck supporting it. To my great fortune computers and virtualization has come along way. Almost all desktop processors have multiple cores and support hardware acceleration ether Intel’s VT-x or AMDV. Memory sizes has grown from 2GB to 8GB or better. Hard disks have gotten ALOT larger to hold all the images needed. Oddly, network speeds have stayed at 1Gb/sec but at least it has gotten cheaper. With all that in mind I should be able to build a test environment to handle quite a bit on a shoestring budget.
Lets take a look at the hardware mix I’ve got.
I started building out machines peace meal like all good home grown labs go. My first cut were Intel Q6600 based machines. Hard to beat them for the quad core goodness and all around performance. All four of these machines have multiple terabytes of hard disk space on hardware raid controllers and 8GB of ram. My second round was also Intel but were smaller using E6750 dual core processors as the base. Fewer hard disks but still 8GB of ram on these two. My third pass I switched gears and went with AMD 810 AM3 based machines. These two also have 8GB of ram and enough space to handle whatever I need them too. Last in this stack is an older AMD 5600+ is used for overflow and testing Windows 2000 on Virtual Server 2005. It has 4GB of ram, still has AMDV and two cores enough to get some work done!
Round them out are a few odd balls and support machines.
P4 3.6Ghz is used as a client load generator.
E6750 with 4GB of ram is used also as a client load generator.
AMD 4800+ acts as my controlling server managing all the images and deployments as well as acting as a Virtual Server 2005 for Windows 2000 as well. 4GB and two cores means its up to the task of functional testing when needed.
You will notice I don’t have any Itanium in the mix. I haven’t had a need for it yet as 100% of all our customers and potential customers aren’t running it. eBay is a quick solution for that though.
That’s the basic hardware setup now moving on to the software stack.
As a startup business I’ve joined the Bizspark and Empower for ISV’s. If you are staring a software or software as service business focused on Microsoft technologies and aren’t using these great programs you are putting yourself at a serious disadvantage. That made the choice of VMware or Hyper-V a no brainer for me. But, just to be fair there were other reasons than cost involved in that decision. I can hear some of you now telling me that VMware ESXi 3.5 is F R E E!!!!! And you would be right. The problem is the hardware it will, or more importantly, won’t run on. Remember this thing is hobbled together and will have a wide mix of hardware that needs to be supported. They also offer VMware for windows (GSX) for free as well. I agree that in general it is superior to Virtual Server 2005 another piece of software comes into play Microsoft Virtual Systems Center. It does manage ESX but I couldn’t figure out if it supported GSX and it does support VS2005 and Hyper-V out of the gate (surprise, surprise) With the hardware stack and now the software stack in place we have to work to get this hodgepodge group of servers installed and configured. As I mentioned before, Hyper-V was pretty much the only choice if I was going to manage it centrally and support my hardware mix. That doesn’t mean it was all plug-n-play so to speak.
Some pros and cons of Hyper-V Server 2008
It supports VT-x and AMDV.
It supports 32bit and 64bit clients.
You can load drivers for pretty much any hardware Windows Server 2008 x64 supports.
It requires x64 support. (Some people really didn’t know)
It requires the disable execution bit to be turned on in bios(not all motherboard manufacturers have the function exposed)
You must have a processor with VT-x or AMDV. If you don’t the OS still installs but the hyper-v services will simply not start.
It is a core installation so no GUI to speak of, if you are a hard core Windows person and has to have the GUI tools you need to load the regular version of Windows 2008 and install the Hyper-V role.
Older clients can run really, really slow. Windows 2000 and Windows XP are prime examples and aren’t listed on the compatibility list there for the need for Virtual Server 2005 R2.
Things to know about Virtual Server 2005 R2. It is also free. Host OS is 32bit support only Windows 2000 or 2003 (no R2). Client wise only up to Windows 2003 32bit is supported but it does run them well enough for what we need.
Installing Hyper-V Sever 2008
The install is pretty simple, it is a Windows 2008 install. If you need to load drivers so the OS can see the disks you can do it with ease even from USB. After the install configuring everything can get tricky simply because you don’t have the GUI to do all the stuff you normally would do post install.
If you aren’t prompted to load your drivers by the hardware detection wizard, aka the hardware was already in the machine, there are some simple and not so simple steps that you can take.
The simplest way:
Locate the drivers you need for ether Windows Server 2008 x64 or Vista x64.
Run the installer from the command prompt.
I found that the Intel INF updater runs out of the box just fine for example.
Slightly harder way:
If the installer won’t run you will need to locate the driver files sys and inf and use the command line pnputil tool to install them.
pnputil –i –a <path to the inf and driver>
Some drivers are rapped in cab files and you may need to extract them before putting them on your server.
run driverquery and locate the driver you just installed, may not be so easy if there is a ton of drivers on the machine.
check to see if the device is now working (nic card, sata/sas controller etc)
If the Hyper-V server is on the network you can use the compmgmt.msc to remotely view services and stuff on the server if the firewall is setup to allow it too.
The HARDEST way:
I found this method on the internet and haven’t verified it but I thought it really showed what some people can do when they think outside the box.
Here’s how I did it, the hard way. But maybe it will help someone else.
1. I built a Lab server, with Windows Deployment Services.
2. Installed Microsoft Deployment Toolkit 2008,
2. Installed the latest AIK Kit: Windows Automated Installation Kit (Windows AIK)
3. Loaded the Hyper-V 2008 Server ISO.
4. Installed all Intel Drivers I will need for old and current builds.
5. Used the Guidelines in Microsoft® Deployment Toolkit 2008
Getting Started Guide, from the Microsoft Deployment Toolkit.
6. Built x64bit Image with LAN Drivers installed.
7. Built a Bootable USB.
8. Booted Server with USB, Did a network install from Windows Deployment Services.
9 Rebooted Server and was able to use NIC Card.
Specific uncommon things to my setup
The next thing on my list is at least one of my installs isn’t joined to the domain. Why would you do that you may ask? Easy, this is the machine that actually hosts the domain controller. I will warn you this isn’t a supported Microsoft configuration. they always recommend having at least one physical domain controller. But since this is a test lab I didn’t want to dedicate a machine to just a DC role. Use a fixed IP, no DHCP if the DC hosts the DHCP you will be stuck with a “chicken and the egg” problem. I personally disable the firewall all together instead of just managing openings since all of these already sit behind a hardware firewall. Issuing this command accomplishes that: netsh advfirewall set allprofiles state off
If the server isn’t a part of the domain managing it requires a few tweaks. John Howard has resolved pretty much all of them in this blog post.
I also ran into an issue where my Windows 7 RC build couldn’t access the hyper-v server and this post fixed that.
Lastly, Sense I am running VS2005 R2 and Hyper-v I have a need to transfer images back and forth these three posts cover that topic as well.
hyper-v to virtual server
As a side effect of setting up this lab and working with SQL Server 2008 and Exchange 2007 SP1 I’m learning how to use Powershell.
As things settle in I’ll be talking more about actually managing the VM’s with System Center and Powershell.
Oh, just a quick shot of the magic closet. Yes, in the lower right hand corner is an AC unit. No, you don’t want to know what my electricity bill is.