In my travels as a Citrix SE I often bring up the topic of “hyper-convergence” in discussions with customers. Some shops are fairly versed on the subject having recently implemented one flavor of it or another in their production environments. Some shops have done little to no investigation of the technology while most are somewhere in the middle and are beginning to take a good look at what it can do for their organizations. Hyper-Converged Infrastructure (HCI) brings together your compute, virtualization and storage systems and integrates them into a single software-defined data center (SDDC) solution. This is a generalized definition, of course. Some commonly used terms to describe HCI: agility, radically simplified, automation, faster deployments, ease-of-management. It doesn’t matter which vendor you’re looking at in the HCI space, they all use some combination of these words to try to help us understand the value their solution provides.
As I spoke with customers about their experiences, I began to wonder how applicable these adjectives really were – after all, who wouldn’t love a solution that was more agile and easier to manage? Larger organizations have various teams dedicated to storage, networking, and server virtualization, all with a wide range of knowledge, skills and formal training where HCI wouldn’t seem as daunting a technology. But what about the smaller- to medium-sized shops, ones that don’t have a large crew or an enviable training budget (or none at all)? What about those engineers who might feel HCI was just another technology heaped onto an already overwhelming plate of responsibility? “Radically simplified” sounds great in a comfy boardroom, but is it valid in the datacenter when you’re trying to get the solution implemented? I mean, how easy can it be when you have to marry storage, virtualization, compute, networking, etc. all together into something you can build upon?
I love the opportunity I’ve been given to participate in Rick Dehlinger’s Project Silverton and have been working on getting myself acquainted with HyperFlex, Cisco’s HCI solution. According to Cisco’s own website http://www.cisco.com/c/en/us/products/hyperconverged-infrastructure/index.html you can get a HyperFlex cluster installed in less than one hour. “Radically simplified,” right? Um, I wasn’t overly confident, to be honest. But I did my research, put in the work, hit the stopwatch, turned the wrenches and learned a ton!
This article will focus on what’s needed to get ready for a HyperFlex installation. I’ll follow it up with Part 2 outlining the install process and Part 3 detailing the post-install tasks. The most important thing to know – it IS possible to deploy HyperFlex in under an hour and we’re going to see how, together.
“Fortune favors the prepared mind.” An individual I highly respect and work closely with uses this quote of Louis Pasteur’s regularly. In the literal sense, it implies a certain amount of luck is required to be successful; on the other hand, it can be interpreted to mean the more prepared you are the less luck you require to become successful. When deploying Cisco HyperFlex, I cannot stress enough the need to prepare, prepare, prepare. And it begins by getting acquainted with the “Cisco HyperFlex Systems Getting Started Guide, Release 1.8”. It’s over 100 pages of information that can quickly get you familiarized with all the pieces of the puzzle. In general, HyperFlex is made up of the following:
- Cisco HyperFlex HX-Series Servers – rack-mount units purpose built as hyperconverged nodes. Both rack-mount and blade servers can be used for additional compute power. Every HyperFlex hyperconverged server ships with the proper version of ESX installed, the Cisco HyperFlex Data Platform Controller VM installed and the basic network configuration completed.
- Cisco HyperFlex Data Platform – a controller VM on each HX node that provides command and control for the software-defined cluster and communicates with the HX Plug-in for vCenter.
- Cisco HyperFlex Data Platform Installer – installs and configures the initial HX cluster, installs the HX Plug-in for vCenter and adds additional HX nodes as needed.
- Cisco UCS Fabric Interconnects (FIs) – provides software-defined compute, storage and network personality leveraging policies via Cisco UCS Manager to ensure consistency throughout the environment.
- VMware vSphere 5.5 or 6.0 – if using HyperFlex 1.8x then upgrading to vSphere 6.0 U2 Patch 3 is recommended.
The guide steps you through all the components, including hardware and software requirements, VLAN configurations, hypervisor and vCenter requirements, etc. Chapter 4 of the guide gets into the physical installation of the gear, how to rack and stack the HX Servers, configuring your Fabric Interconnects and connecting the servers to the FIs. Chapter 5 details the steps required to actually install HyperFlex, and it starts by deploying the HX Data Platform Installer 1.8.1c on an existing ESX host, VMware Workstation or VMware Fusion. It’s important to know you cannot deploy the HX Data Platform Installer on a ESX host that will be part of the HX cluster, but more on this later. The remaining chapters detail post-installation processes and additional configurations like adding multiple clusters or additional HX Servers to an existing environment.
Once everything was physically installed and cabled, I simply connected via console cable to the first FI and entered Setup Mode (it gives you an option to install from scratch or restore from a backup). The config process is very straightforward: set the admin password, physical IP address, FI Cluster IP address, subnet mask, default gateway and DNS settings. Now console into the second FI, enter Setup Mode and answer “Yes” when asked if this FI is part of a cluster. Once the FI Cluster has initiated, open a browser and navigate to the Cluster IP address. Login with the admin credentials and verify the firmware version. Cisco is currently shipping 3.x version, so navigate to Equipment -> Firmware Management -> Installed Firmware to verify 3.1(2b) is the installed infrastructure firmware.
The next document you’ll want to have is the “Cisco HyperFlex Systems HX Data Platform Pre-Install Checklist.” This checklist is extremely helpful and will save you so much time. By having all your environment details documented it makes it extremely easy to be able to punch-in the proper values during the installation process. At only 15 total pages it lays out what you need to have in place prior to attempting an installation. This document can also be completed and given to your Cisco account team to create a configuration file (JSON file) that can be used for the initial setup. Installing via JSON configuration file would eliminate the need to manually input almost all values, significantly decreasing both the deployment time and the risk associated with human error. Some snippets of the checklist:
There were just two more things left to do before I could actually begin the HyperFlex installation – download some firmware and configure the FI ports.
UCS Manager 3.1(2b)A firmware was on the FIs, but two additional bundles needed to be downloaded. I needed firmware bundle 3.1(2b)C for the C-Series (chassis) HX Servers and 3.1(2b)B for the B-Series (blade) HX Servers. It’s important to know you still need the B-Series firmware bundle even if you only have C-Series servers in your cluster. If you do not have both firmware bundles downloaded you will get an error during the HyperFlex installation process. The firmware packages can be downloaded from Cisco and uploading them to the FIs was easy. Simply fire up a browser and navigate to the FI Cluster IP and launch UCS Manager. Then navigate to the Firmware Management tab and you’ll be able to select the “Download Firmware” button. Browse to your local computer to where you saved the firmware bundles and it will then upload them to UCS Manager:
Configuring the FI ports was also not difficult. While in UCS Manager expand the Fabric Interconnect node on the left, select one FI and expand it, then click on “Ethernet Ports.” On the right you’ll select the ports for your HX Servers, right-click and choose “Configure as Server Ports.” You’ll then select the other port that’s to be used by the FIs as their uplink port, right-click and choose “Configure as Uplink Port.” It’s important to know you must configure the Server and Uplink ports on both FIs. When you’re done, it will look something like this:
And if you’re seeing these screenshots and thinking, “…my UCS Manager doesn’t look anything like that…” you’d be right if you were not running UCS Manager 3.x. Here’s what the HTML version of UCS Manager 3.1(2b) default view looks like:
For anyone used to UCS Manager 2.x it’s a significant UI change. Personally, I find it more intuitive and easier to use. I would also recommend using Internet Explorer instead of Chrome to login to UCS Manager, the UI seemed much more consistent and responsive with IE.
So let’s recap. My HyperFlex gear is racked, stacked and cabled. VLANs and routing figured out, my FI UCS Domain created. The FI Server Ports and Uplink Port have been set. I’ve downloaded both firmware bundles 3.1(2b)C and 3.1(2b)B from Cisco and uploaded them in UCS Manager. I’ve completed my Pre-Install Checklist and have everything documented. Awesome – I should be ready to actually start the installation of HyperFlex. My next article in this blog series will cover the HyperFlex install process and lessons learned; in the meantime, below are links to the documents I highlighted today.
Cisco HyperFlex Systems Getting Started Guide, Release 1.8
Cisco HyperFlex Systems Data Platform Guide, Release 1.8
Cisco HyperFlex Systems HX Data Platform Pre-Install Checklist