Rename Graviton section to "Taints and Tolerations"

[rizblie]

What would you like to be added?

The section in Fundamentals -> Managed Node Groups -> Graviton Instances should be renamed to reflect that it is really an lab on Taints and Tolerations. The fact that it is using Graviton is just an example, and the same technique could be used for other purpose e.g. GPU instances.

Secifically referring to: https://github.com/aws-samples/eks-workshop-v2/blob/main/website/docs/fundamentals/managed-node-groups/graviton/index.md

Why is this needed?

This is first and foremost an EKS workshop , so the audience interest is in Kubernetes and how to run Kubernetes on AWS. The general concept that the audience wants to learn is how they can use taints and tolerations with managed node groups, not Graviton specifically. So the current title is not helpful.

[svennam92]

Hey @rizblie, thanks for opening this issue. This section was originally titled "Taints and Tolerations" before we made modifications to focus more on Graviton. The original module wasn't in line with our workshop tenets — in general our modules are built with a focus on AWS/EKS-related concepts. See Service Differentiation in our tenets.md.

[rizblie]

Hi - I can see that the module "Graviton Instances" is more than being just about taints & tolerations, because it also introduces the concept of multi-architecture images. However, a key part of this module is about how you can use Managed Node Groups with taints in order to facilitate placement of pods on specific hardware types. Whether it is Graviton, GPU, or some other specific instance type is just a specific example of how Managed Node Group taints can be used.

With reference to the tenets for the workshop, note that Graviton has nothing to do with EKS. You could use K8s with Graviton instances without using EKS.

If you see it from the point of view of an SA delivering this workshop to a customer... I had a case recently where a customer wanted to understand how they could use GPU for some pods in their EKS cluster. Confusingly, I had to ask the customer to look at the "Graviton Instances" module!

There is no easy answer. One option is to have a separate module on "Managed Node Group taints for compute placement.". And then a separate "Graviton" module which places more emphasis on the multi-architecture container image concept and the benefits of using Graviton with EKS.

[niallthomson]

For me the narrative of this lab is "Using Graviton with EKS Managed Node Groups", which I think demonstrates some valuable concepts such as how to configure them (specifying appropriate AMI type), the fact that x86 and arm64 nodes can exist in the same cluster, and how to differentiate between them for scheduling purposes if needed.

We're discussing how we can add more content about more concepts that are core to Kubernetes and less specific to EKS, which would cover the taints portion, but thats still in the early phases.

It might be worth us taking a step back on this lab as it exists and figure out where we can add more content that makes the focus less about taints. For example adding in the eks-node-viewer to show cost savings.

[rizblie]

Agreed @niallthomson.

However, with regard to taints I think we need to distinguish between the concept of K8s taints (which is nothing to do with EKS), and the concept of Managed Node Group taints which is a feature of EKS i.e. the fact that you can update taints on Managed Node Groups via the EKS API. So I would not place this in the "less specific to EKS" category.