π’ CloudStack Kubernetes Flavors β A Reference Guide I Wish I Had When I Started #13638
chunkyen
started this conversation in
Show and tell
Replies: 1 comment 1 reply
-
|
Dear @chunkyen, I am personally interested in what you are doing with Talos inside ACS. Are you working on alternative offering for Kubernetes using Calico instead of the official K8s? If that is where you heading, I am available to perform some tests of the Calico build on ACS because that is my preferred way of running K8s anywhere. Feel free to ping me if you need assistance on that. Regards, |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
I've been exploring Kubernetes on CloudStack out of personal interest, and one thing that struck me early on was how scattered the information is. Documentation exists for each individual approach (CKS, CAPC, Talos), but there wasn't a single place that compared them side-by-side or walked through the practical gotchas you hit along the way.
So I started building that reference β a repository covering four approaches to running Kubernetes on CloudStack:
1οΈβ£ CKS (CloudStack Kubernetes Service) β Native CloudStack integration
2οΈβ£ CAPC (Cluster API Provider for CloudStack) β Infrastructure-as-Code with CAPI
3οΈβ£ Rancher + CAPC β Managed Kubernetes with Rancher as the management plane
4οΈβ£ Talos Linux β Minimal, immutable Kubernetes OS (standalone or with Sidero Omni)
Plus two cross-cutting components that apply to every flavor:
Each flavor includes architecture analysis, setup guides, and β most importantly β the lessons learned from actually deploying them. Things like the CSI DaemonSet cloud-init-dir issue on Talos, the ClusterClass limitation with CAPC, or the self-signed CA trust dance with Omni. The kind of details that take hours to figure out but are obvious in hindsight.
A significant portion of the work went into air-gapped / offline CKS deployment β including a deep-dive into the ISO build process, container image pre-import workarounds, Cilium manifest patching for offline environments, and the various failure modes you hit when CKS assumes internet access is available. The offline guide (https://github.com/chunkyen/cloudstack-kubernetes-flavors/blob/main/setup/cks/cks-offline.md) documents all of this.
The Rancher + Turtles + CAPC combination was something I figured out on my own β I couldn't find any existing architecture or guide that documented this specific stack working together. It's a good example of how open source technologies can be combined even when there's no official reference architecture, and I hope the guide helps others who want to run this stack without having to piece it together from scratch.
I've also included suggested improvements for the approaches that could benefit from upstream changes:
What's in the repo:
How it was built: The entire repository was developed collaboratively with AI agents (Hermes / OpenCLAW) running on locally hosted models (Qwen, Gemma, etc.) β a testament to how far open-source LLMs have come for technical documentation and infrastructure work.
π https://github.com/chunkyen/cloudstack-kubernetes-flavors
This is a lab project driven by curiosity, not production experience. If you're also exploring Kubernetes on CloudStack, I hope this saves you some of the head-scratching. Corrections, contributions, and suggestions are very welcome!
Beta Was this translation helpful? Give feedback.
All reactions