How To Upgrade Hyperconverged Infrastructure
So your Hyper-Converged Infrastructure (HCI) cluster is not performing the way it used to. You have run short of something; storage, maybe. Perhaps you are low on processing power, or network bandwidth. It’s upgrade time, but what kind of experience can you expect with HCI? How is it different from upgrades to traditional storage and compute?
Upgrade In Place
Though upgrading in place is the less common method of upgrading an HCI appliance, HCI vendor that allow field upgrades tend to focus on minimizing the component-swapping and fiddling with screwdrivers. For some organizations, upgrading your existing units without adding any nodes to the cluster is the way to go, and this is easier than you might assume.
Upgrading in place involves swapping out pieces of your appliances rather than buying new nodes. This is where you might add more drives, upgrade to larger drives, or go for a different mix of flash and HDDs. On some HCI appliances, Network Interface Cards (NICs) can also be swapped to increase network throughput.
This type of upgrade may offer a lower capital outlay than buying additional nodes, or replacing an entire cluster. After all, organisations are only buying new parts, not a whole new system.
Those looking to embark on this journey should, however, bear in mind that upgrading in place may be costly in terms of service interruption. HCI nodes undergoing upgrade need to be shut down and taken apart for in-place upgrades, and with many vendors it means taking the whole cluster off-line, not just one node at a time.
Organizations who feel that upgrading in place is likely to be part of their scaling strategy should engage with their vendor if they are considering this kind of upgrade, and make sure that what they want to do is supported. For organizations seeking more of a performance or capacity improvement than can be achieved by upgrading parts – or who have ease of use or uptime concerns – then the rolling upgrade is the route to consider.
Rolling Upgrade
An HCI rolling upgrade begins with determining what class of nodes need to be added. Decisions are made about whether the required additional nodes should be storage-heavy or compute-heavy. Here, customers should focus on current, short-term projected needs. Don’t fret too much about future-proofing, because one can add more nodes later, so there’s less of a need to overprovision, r to overspend.
Customers should also investigate whether or not the vendor restricts what types of nodes can go in the same cluster. Some do, but the better vendors let you mix and match to varying extents. If you can’t mix and match, upgrades get a lot more complicated, and it’s best to know a vendor’s policy on this before starting one’s HCI journey with them. It’s also best to know all of the above before one engages with salesbodies.
Once the nodes are selected that will meet requirements the rest is comparatively simple. With HCI there’s just one vendor, so there’s just one sales team to deal with. More importantly, with HCI there should only be one support team to deal with. In industry parlance this is called “having a single throat to choak”, and it is generally a good thing.
Once the new nodes are purchased, it gets shipped. There shouldn’t be any waiting on multiple components from multiple vendors. Also generally a good thing.
Once unboxed and racked, the nodes need to be incorporated into the existing cluster. Node incorporation can be done without needing to shut down production workloads or disrupt users.
The vendor may have sent instructions, or may have a support person standing by to walk customers through setup. For most solutions, cluster incorporation involves some simple commands in the management interface to tell the cluster to absorb a new node and migrate workloads accordingly. Depending on vendor support, a technician may be manage this process remotely.
Adding a node to an existing cluster can take as little as 5 minutes from power on to serving workloads. This simplicity is why rolling upgrades are regularly considered one of the more attractive features of HCI.
A Refreshing Change
HCI can scale both up and out. Capable vendors ensure customers have the option of both upgrading existing HCI nodes as well as adding new ones. In some cases it is common to upgrade an existing cluster’s nodes while adding new ones.
One example of this might be if a customer were to buy an HCI cluster that had all mechanical hard disks, but eventually decided they required both more performance and more capacity. They might decide to purchase additional nodes for their existing cluster, but have those new nodes by hybrid nodes. In order to keep the cluster’s capabilities balanced, they may also decide to upgrade the existing nodes with flash drives and faster NICs.
This flexibility drives HCI adoption. The primary selling point of HCI remains the stark contrast it offers to the forklift upgrade. Forklift upgrades require organizations to rip and replace most of their infrastructure in one pricy, disruptive effort. HCI’s lifecycle, meanwhile, unfolds a little differently.
Fear not the HCI upgrade. The HCI upgrade is good.