Migrate Kubernetes cluster autoscaler to Karpenter

We will migrate a Amazon EKS Kubernetes cluster using the cluster autoscaler to Karpenter for node autoscaling.
Check out the full documentation here karpenter.sh/v0.9.0/getting-s...
This video explains every step in the documentation and will show you what happens when you run the commands and how to verify Karpenter is working.
#aws #kubernetes #karpenter
00:00 - Intro
00:20 - EKS Cluster
00:48 - Karpenter docs
15:09 - Outro

[@hawking8987]

great demo ，thanks for sharing

[@techmiker]

@ 5:26 -- How did you copy to the clipboard there??? 😄

[@gunasekhar1102]

How does cost reduces if we use karpenter, for ex 1 m series 4 cpu 16 gb costs more than the 2 t series 2 cpu and 8 gb ram

[@JustinGarrison]

Karpenter will pick the best nodes for your workloads based on the workload requests. If you want spot or a specific instance type you can define it in your workload manifest. A t2 instance has burstable performance that is not guaranteed. If your workload can work with burstable performance you can define that but not all workloads can.

[@icurtispe]

Installation on a running EKS Cluster with no NodeGroups, several Security Groups, and taints is very complicated. I wish installation was more straight forward. We try KubeCost add-on and it was easier.

[@randomfandom53]

How do I get rid of the final two nodes to make karpenter control everything? Just delete them?

[@ContainersfromtheCouch]

Karpenter has to run somewhere and shouldn’t run on nodes it provisions. You either need to keep nodes for Karpenter and other cluster services or run it on fargate

[@2mahender]

why is karpenter preferred, over CAS?

[@ContainersfromtheCouch]

CAS is great but ends up being a lot to manage for clusters with lots of different workloads. Karpenter takes away predefined node groups and adapts to workload needs as defined

[@Luther_Luffeigh]

Sorry too soon; that thing isn’t ready

[@user-zv1mu9lt8c]

Karpenter hasn't seemed great as of late