i assume when people are talking about k8s complexity, it’s either more complicated scenarios, or they’re not talking about managed k8s.
even then though, it’s more that complex needs are complex and not so much that k8s is the thing driving the complexity.
if your primary complexity is k8s you either are doing it wrong or chose the wrong tool.
That’s fair! Nutanix impressed me as well when I was doing a hypervisor deep dive in 2022/2023, but I had concerns about their (lack of) profitability in the long run. VMware Tanzu wasn’t bad either, but was more of an arm-pull than AWS was for K8s. Talos is on my “to review” list, especially with their community license that let’s you manage small deployments like a proper Enterprise might (great evangelism idea, there), but moving everything to kube-virt was a nonstarter in the org at the time.
K8s’ ecosystem is improving by the day, but I’m still leaning towards a managed K8s cluster from a cloud provider for most production workloads, as it really is just a few lines of YAML to bootstrap new clusters with automated backups and secrets management nowadays - if you don’t mind the eye-watering bill that comes every month for said convenience.
If you work for any of the CISA-Identified 16 critical infrastructure sectors, one of their recommendations is for organizations to be able to expect to operate for more than 24h without an Internet connection.
Kinda hard to control real-world things with no Internet connection that rely on an internet connection
Note: Nutanix made some interesting k8s-related acquisitions in the last few years. If interested, you should take a look at some of the things they are working on.
If I were still in that role, I’d absolutely be keeping my Nutanix rep warm for a possible migration. Alas, I’m in another org building them a Win11 imaging pipeline for the time being, and Nutanix doesn’t want to play nice with my personal N100 NUCs for me to try their Community Edition.
> or they’re not talking about managed k8s
Bingo! Managed K8s on a hyperscaler is easy mode, and a godsend. I’m speaking from the cluster admin and bare metal perspectives, where it’s a frustrating exercise in micromanaging all these additional abstraction layers just to get the basic “managed” K8s functions in a reliable state.
If you’re using managed K8s, then don’t @ me about “It’S nOt CoMpLeX” because we’re not even in the same book, let alone the same chapter. Hypervisors can deploy to bare metal and shared storage without much in the way of additional configuration, but K8s requires defining PVs, storage classes, network layers, local DNS, local firewalls and routers, etc, most of which it does not want to play nicely with pre-1.20 out of the box. It’s gotten better these past two years for sure, but it’s still not as plug-and-play as something like ESXi+vSphere/RHEL+Cockpit/PVE, and that’s a damn shame.
Hence why I’m always eager to drive something like Canine!
(EDIT: and unless you absolutely have a reason to do bare metal self-hosted K8s from binaries you should absolutely be on a managed K8s cluster provider of some sort. Seriously, the headaches aren’t worth the cost savings for any org of size)