Azure Stack with Only Two Servers

Ok, so that title is a teeny bit of an overstatement. What we’re going to discuss today is not the Azure Stack you have heard so much about. No, today we’re talking about an Azure Stack HCI Cluster.

Essentially, it’s a Storage Spaces Direct cluster with brand new shiny marketing and management. While you can use the old “*.msc” tools, it’s a far, far better thing to use the new Windows Administrator Center (WAC)! More on that soon.

I’m not going to dig into the details of how S2D works or the fine details of building clusters large and small. Instead, I want to share with you some of the reasons why small and medium businesses need to pay attention to this technology pronto.

  1. Scalability: Sure, most of the documentation you find today focuses on building clusters of 4-6 nodes or more. That’s great unless you are a small business that doesn’t need that kind of compute and storage. That’s where the two-node “back to back” cluster comes in. Think “entry-level”. The beauty of this solution is scalability. If you do outgrow this, you buy a new node, possibly a new network switch and you are up to a 3-node cluster!
  2. Compact: I had two old servers that took up two Rack Units (RU) of space each plus an external drive array that took another 2 RU. That totaled up to 6 Rack Units or 10.5 inches of rack space. The new servers that replaced them are 1 Rack Unit each for a total of 3.5 inches! That doesn’t even touch on noise, heat and power consumption.
  3. Fast and easy: Ok yes, it is a complicated technology. However, you can follow Dell’s Deployment Guide for 2-Node hyper-converged back-to-back connected infrastructure with R640 Storage Spaces Direct Ready Node
    and you’ll be 90% of the way there. It’s a mouthful, but it’s 40 pages of step by step instructions and checklists. *TIP* I’ve included some tips below for a couple of places that aren’t super clear (or at least to me)
  4. Well Documented: If you are like me and want to understand how this all works before you bet your mission critical VMs to it, there is a ton of good information out there. Here are some options depending on your preferred method.

    • The Book: Dave Kawula’sMaster Storage Spaces Direct.  It’s over 300 pages of detailed explanation for a very reasonable price. Although you can get it free, those people spent a lot of time working on that so pay something. You’ll see what I mean on the Leanpub site.
    • The Video: Pluralsight’s Greg Shields has a course on Storage Spaces Direct that is 4 hours of in-depth instruction on Windows Failover Clusters and Storage Spaces Direct in Windows 2019. If you aren’t a subscriber to Pluralsight, they offer trial periods!
    • The ‘Official’ Documentation: Microsoft’s Storage Spaces Direct Overview is the place for the official documentation.

There are a few tips and gotchas that I want to share from my experience.
First, the hardware. These servers aren’t crippling expensive, but they certainly aren’t disposable cheap either. This means there are a lot of critical hardware decisions to make. What’s more important to your organization – budget or speed or a balance? On the one end, you can go all-flash storage which will give you a fast system, but the price tag goes up fast too. The less expensive but slightly more complicated set up is a hybrid of SSD and HDD storage. Making certain that you have the right mix of memory, storage and network adapters can be a daunting task.

Honing a specification and shopping it around to get the perfect balance of form, function, and price is great if you are a hobbyist or plan on building out a ton of these at one time.

However, for IT Admins in most small companies, the more important thing is that it gets done quick, quiet and correctly. The business owners don’t care about the fine nuances. They want to start realizing the business benefits of the new technology.

I chose a much more economical option, both in time and cash. I searched Dell’s website for “Microsoft Storage Spaces Direct Ready Nodes” and picked out a pair of their “Ready Nodes” that looked like they matched my budget and needs.

Then it was a small matter of reaching out to my Dell team. My sales rep put me in touch with their server/storage specialist. He asked a couple of questions about workloads, storage, and networking. Presto, we had our servers on order.

*Pro-tip* Buy the fiber patch cables at the same time. They are no less expensive elsewhere and you have less chance of getting crap cables.

If you don’t already have a relationship with Dell, there are several other Microsoft certified hardware vendors. There is a list here: http://bit.ly/S2D2NodeOptimized

 Tips for the build

You’ve only got 2 servers to configure, so typing each command is feasible. However, in the interest of both documenting what I did and saving on silly typos, I opened up VSCode and wrote up all the Powershell commands described in the Dell docs.

The steps (much simplified) look something like:

  1. Install OS if not preinstalled and then patch to current.
  2. Plan and Map out IP networking. Use the checklists in the back of the Dell guide to map out your IP scheme. It’s a time saver!
  3. Pre-stage the Active Directory accounts for the Cluster as documented here. Trust me, it’s faster to do it on the front side than after the fact.
  4. Install Windows Features on each node
  5. Virtual Switch and Network configuration on each node
  6. Test-Cluster and Review/Remediation
  7. Create Cluster – with no storage
  8. Enable Storage Spaces Direct on the cluster.
  9. Configure a witness – either another server on the site or a cloud witness.
  10. Create virtual disks – Cluster creation and enabling Storage Spaces Direct on the cluster creates only a storage pool and does not provision any virtual disks in the storage pool.

Next time, we’ll go into depth on the management of our newly built Azure Stack HCI setup!