❗ ARCHIVED ❗

With #1210 being merged, kubernetes-sigs/descheduler now respects preferredDuringSchedulingIgnoredDuringExecution term, so there's no need to use and maintain this program

kube-better-node

This small program does the only thing kubernetes-sigs/descheduler can't do. It will evict Pods which could be potentially schedulled to a better Node, based on preferredDuringSchedulingIgnoredDuringExecution nodeAffinity terms.

⚠️ Disclaimer ⚠️

• I have no intention to replicate and follow all Descheduler policies and rules, but feel free to create a PR if something is missing.
• I do not need functionality provided by this program anymore, however I still see it useful for exploring the Golang world.
• Remember, it' a workaround and there's no guarantee that a Pod will be scheduled to a Node you expect it to be.

Why?

There're at least two rotten unmerged PRs in the Descheduler repository:

• #130 (FindBetterPreferredNode and CalcPodPriorityScore are inspired by it. Thanks @tsu1980)
• #129

In fact, this functionallity will be never implemented in Descheduler.

As @asnkh said in #211:

...unless descheduler can make the same decision as kube-scheduler, it can cause this kind of ineffective pod evictions. I have no good idea to overcome this difficulty. Copying all scheduling policies in kube-scheduler to descheduler is not realistic.

Use case

However, there're real world use-cases, where such dumb eviction can be usefull.

For example, when your project is small and there's a need to use some special Spot instance node type (such as Tesla p2.xlarge). This type of instance can be taken back from you by AWS, but if your project can afford some performance degradation for a short period of time (untill there's a new node of p2.xlarge given), kube-scheduler will place Pods on some other Node available at the moment.

But when finally new Tesla p2.xlarge node will apper and join the cluster, there's nothing to schedule your Pods back to it.

Here kube-better-node can be usefull.

Usage

❯ kube-better-node -h
Usage of kube-better-node:
  -dry-run
    	Dry run
  -tolerance int
    	Ignore certain weight difference
  -v value
    	number for the log level verbosity

Installation

helm repo add decayofmind https://decayofmind.github.io/charts/
helm install kube-better-node decayofmind/kube-better-node

Links

• https://github.com/kubernetes-sigs/cluster-capacity