I'm currently using AWS EKS Hybrid nodes alongside three on-premise NVIDIA GPU nodes that we've already set up. However, we're looking to migrate away from EKS because it's turning out to be about 80% more expensive than we anticipated. We're leaning towards using RKE2 and are also considering Talos Linux. Does anyone have suggestions or insights on the best way to handle Kubernetes distributions for hybrid setups, especially ones that can handle LLM and other GPU-intensive workloads?
3 Answers
We ended up going with RKE2 and Rancher to standardize our RBAC across different clusters, regardless of where they are located. Just be cautious; it sounds like you might need Rancher Prime for comprehensive management. We saw a big price hike when SUSE changed our licensing requirements to keep us updated with Harvester.
I think Talos is a solid choice! It provides a streamlined experience, and if you're looking at a multi-cluster setup, you might want to check out Omni for managing everything. However, RKE2 and Rancher are also good options but keep in mind they seem to fit more within cloud environments.
In addition to the technical aspects of node-to-node networking, I suggest looking for a distribution that uses Konnectivity. It simplifies the API server and Kubelet communication immensely. If you're considering multiple clusters, Kamaji could work great for your setup. It's ideal for having control planes in the cloud and nodes on-prem, or even the other way around. If you stick to a single cluster, k0s could be worth a look.
Related Questions
How To: Running Codex CLI on Windows with Azure OpenAI
Set Wordpress Featured Image Using Javascript
How To Fix PHP Random Being The Same
Why no WebP Support with Wordpress
Replace Wordpress Cron With Linux Cron
Customize Yoast Canonical URL Programmatically