Why is Autopilot Kubernetes Cluster is so expensive compared to standard cluster on GKE?

Viewed 754

I run software of a website on Google Kubernetes Engine. I wanted to try autopilot clusters rather than standart clusters. After days of trial, I saw that it is so expensive. Before autopilot clusters, I used standart clusters with 2 nodes (2 cpu, 4GB ram for each node, medium one). Here is the chart from billing. Right hand side is the cost of autopilot, the top blue area is autopilot cpu cost.

Do you have any explanation? It seems that autopilot clusters are expensive, do you think it justifies the cost with its features?

enter image description here

1 Answers

When you create a pod in kubernetes autopilot you have to specify the resources that the pod will consume (by default are 0.5 cpu, 1GiB of ephemeral storage and 2 GiB of RAM). Check the resources that you are requesting and adjust them to your needs (you can use an autoscaler or manage them manually).

Also keep in mind that autopilot is more expensive than default GKE (and the minimum resources requested by a pod are higher too)

Autopilot resource ranges

Related