Building Stability: Pitfalls of Not Setting Kubernetes Resource Requests and Limit

Managing container resources is a Kubernetes basic concept that requires users to set resource requests and limits. It can be a difficult concept to grasp because of the ways in which the scheduler, OOM killer, cluster autoscaler and horizontal pod autoscaler interact. Not setting requests/limits correctly can cause a variety of issues such as excessive cost, instability and outages.

In this talk, Andy Suderman of Fairwinds will walk you through how resource requests and limits work and, through four live demos, demonstrate what happens if they are misconfigured. During the live demos, he’ll show Prometheus metrics so you can see firsthand what’s happening in the cluster. You’ll leave with an understanding of why and how to configure resource requests and limits to build a stable production environment.

[AnkitKumar-yiir]

Let suppose we are in prod server and suddenly lots of load comeup on the server. We have autoscale configure at the node but still it take time around 50-60sec to spin up but until new nodes come up, my pods shows error as evicted pod error.
How to tackle this kind of sitation ?