Learn Kubernetes Weekly issue 168 · 28 Jan 2026

Deleting CRDs, Carbon-Aware ML Training, vLLM Setups for Spiky Traffic, Argo Resource Protection, Reproducible Infrastructure with NixOS

This newsletter is brought to you by Kubex — Automated Resource Optimization for Kubernetes, GPUs and AI Workloads

Articles

  1. What Happens When You Delete a Kubernetes CustomResourceDefinition?

    marcincuber.medium.com

    This article explains what happens internally when you delete a Kubernetes CRD, covering finalizers, cascading deletion, and orphaned custom resources.

  2. Kubernetes Resource Management ebook for Platform Owners

    kubex.ai

    Platform teams face a resource paradox: clusters full of unused capacity while developers demand more.

    This guide shows how to manage Kubernetes resources at scale without slowing teams down or wasting budget.

    sponsored

  3. Making ML Training Carbon-Aware with Compute Gardener

    medium.com

    This article describes how Compute Gardener, an open-source Kubernetes scheduler plugin, automatically delays non-critical ML training jobs until electricity grid carbon intensity is lower to reduce carbon emissions without modifying application code.

  4. 8 vLLM Serving Setups That Handle Spiky Traffic

    medium.com

    This article compares 8 different vLLM serving setups for handling spiky traffic in production LLM deployments, including autoscaling strategies and architecture patterns.

  5. How I Prevent My Kubernetes Resources from Being Deleted When Argo Apps Are Removed

    mirilittleme.medium.com

    This article explains how to prevent Kubernetes resources from being deleted when Argo CD Applications or ApplicationSets are removed.

  6. Reproducible Kubernetes infrastructure with NixOS and OKD

    medium.com

    This article demonstrates how to build reproducible Kubernetes infrastructure using NixOS and OKD for declarative node configuration and cluster management.

Articles worth checking out:

Right-Sizing GPUs in Kubernetes: A Platform Engineer’s Guide to AI Infrastructure Efficiency

GPU spend is rising faster than the tooling to manage it.

Most teams show healthy utilization while requests overshoot real usage by 3x. This ebook shows how to fix GPU waste before costs spiral.

Get notified at launch

Right-Sizing GPUs in Kubernetes: A Platform Engineer’s Guide to AI Infrastructure Efficiency

Tutorials

  1. Getting my feet wet with Crossplane

    blog.frankel.ch

    This tutorial teaches how to use Crossplane on a Kubernetes cluster to provision, configure, and manage a Google Kubernetes Engine (GKE) cluster via declarative YAML manifests.

  2. Deploying an OpenTelemetry-Native LGTM Stack on Kubernetes

    awsmorocco.com

    This tutorial teaches how to deploy a complete LGTM observability stack (Loki, Grafana, Tempo, Mimir) with OpenTelemetry on Kubernetes for unified metrics, logs, and traces.

  3. How we reduced a container image by 99.7% (800gb to 2gb) | sealos blog

    sealos.io

    This case study shows how Sealos reduced a container image from 1.2GB to 100MB by:

    • optimizing the base image,
    • using multi-stage builds,
    • and removing unnecessary dependencies in a Next.js application.
Running a Full Kubernetes Cluster for $2 a Month

Most developers assume Kubernetes requires an enterprise budget. Varnit Goyal proves otherwise — he built a full three-node Kubernetes cluster for $2.16/month using Rackspace Spot Instances.

The trick: pick non-default instance types, distribute nodes across low-demand regions, and let Kubernetes handle rescheduling when nodes get preempted. For service exposure, he replaced the $10/month load balancer with Tailscale Funnel — free.

In this episode:

  • How Spot Instance bidding works and which strategies keep costs and preemption low
  • Using Tailscale Kubernetes operator as a free alternative to traditional load balancers
  • Running real development dependencies (Kafka, Elasticsearch, Postgres) on a budget cluster

A practical walkthrough of what Kubernetes actually needs to function — and what you can strip away.

Running a Full Kubernetes Cluster for $2 a Month

Kubernetes jobs

    • Platform Engineer with AMERICAN SYSTEMS

    • Salary: $155.7K to $260K a year

    • Location: remote from the United States of America

    • Tech stack: Kubernetes, Ansible, Shell, Python, Gitlab, Helm, Rancher

    • Systems Performance Engineer with Datadog

    • Salary: $187K to $240K a year

    • Location: based in the office (and remote from home) in Boston, MA / New York, NY, USA

    • Tech stack: Kubernetes, Datadog, containerd, AWS, Go, Javascript, Rust, C

    • Platform Engineer with Peraton

    • Salary: $112K to $179K a year

    • Location: based in the office in Washington, D.C, USA

    • Tech stack: Kubernetes, Grafana, AWS, Terraform, Shell, Python, Prometheus, Helm, ArgoCD

    • DevSecOps Engineer with Corelight

    • Salary: $221K to $268K a year

    • Location: remote from North America

    • Tech stack: Kubernetes, Go, Splunk

    • DevOps Engineer with Corelight

    • Salary: $158K to $198K a year

    • Location: remote from North America

    • Tech stack: Kubernetes, Grafana, AWS, Azure, GCP, Docker, Terraform, Pulumi, Ansible, Puppet

Discover more Kubernetes jobs on Kube Careers →

Code & tools

  1. ArgoCD-agent: scalable GitOps

    github.com/argoproj-labs

    This tool extends Argo CD with a hub-and-spoke agent architecture in which lightweight agents in remote clusters connect back to a central control plane.

  2. Descheduler: Kubernetes pod balancer

    github.com/kubernetes-sigs

    You can use the Kubernetes Descheduler to evict pods based on specific strategies so that the pods can be rescheduled onto more appropriate nodes.

  3. ForkSpacer: environment forker

    github.com/forkspacer

    ForkSpacer is a Kubernetes tool that spreads pods across nodes based on resource usage and topology to prevent hot spots and reduce noisy neighbor problems.

  4. Lynq: SQL-like Kubernetes

    github.com/kubernetes-tenants

    Lynq is a Kubernetes-native tool that lets you query data inside your cluster using SQL-like language and view results from pods, nodes, and services without writing custom code.

  5. Headlamp: Extensible Kubernetes Web UI

    github.com/headlamp-k8s

    Headlamp is an easy-to-use and extensible Kubernetes web UI providing a user-friendly interface for managing and debugging Kubernetes clusters.

Other interesting projects:

Subscribe to Learn Kubernetes Weekly

Trusted by 77K engineers. Delivered 168 issues and counting.

or subscribe via

Upcoming Kubernetes events

  1. Jan

    29

    Advanced Kubernetes course

    Online workshop organized by LearnKube.

    • This is a virtual event

    • This event requires an entrance fee

  2. Jan

    29

    Eliminate OOM Kills and Resource Waste with Automation

    Online webinar organized by Kubex.

    • This is a virtual event

    • This is a free event.

  3. Jan

    29

    GitOps on Day Two, or How to Bring Order to Chaos with Harbor, ArgoCD, Renovate, and N8N

    In-person meetup organized by Cloud Native Sofia.

    • Location: Sofia, BG

    • This is a free event.

  4. Feb

    1

    FOSDEM 2026

    In-person conference organized by FOSDEM.

    • Location: Brussels, BE

    • This event requires an entrance fee

  5. Feb

    3

    Cloud Native Days France 2026

    In-person conference organized by CND France.

    • Location: Paris, FR

    • This event requires an entrance fee

  6. Feb

    4

    CfgMgmtCamp 2026 Ghent

    In-person conference organized by CfgMgmtCamp.

    • Location: Ghent, BE

    • This event requires an entrance fee

  7. Feb

    12

    Container Days London

    In-person conference organized by Container Days.

    • Location: London, UK

    • This event requires an entrance fee

      • Use KUBEEVENTS to get 50% off

Discover more Kubernetes events on Kube Events →

Thanks to our sponsors who make Kube Today possible

  • LearnKube
  • Akamai
  • Fairwinds
  • Densify
Find out more about being a sponsor →

Kubernetes call for papers

  1. 17

    days

    CfgMgmtCamp 2026 Ghent

    The Call For Paper is open until 15 February 2026 at UTC. More info →
    • Location: Ghent, BE

    • In-person conference organized by CfgMgmtCamp.

    • The conference starts on the 4 February 2026.

    • Apply here
  2. 30

    days

    Container Days Hamburg 2026

    The Call For Paper is open until 28 February 2026 at UTC. More info →
    • Location: Hamburg, DE

    • In-person conference organized by Container Days.

    • The conference starts on the 4 September 2026.

    • Apply here
  3. 36

    days

    Cloud Native Days Italy 2026

    The Call For Paper is open until 6 March 2026 at UTC. More info →
    • Location: Bologna, IT

    • In-person conference organized by CND Italy.

    • The conference starts on the 18 May 2026.

    • Apply here
  4. 3

    days

    KubeCon + CloudNativeCon India 2026

    The Call For Paper is open until 2 February 2026 at UTC. More info →
    • Location: Mumbai, IN

    • In-person conference organized by CNCF.

    • The conference starts on the 19 June 2026.

    • Apply here
  5. 26

    days

    Kubernetes Community Days Beijing 2026

    The Call For Paper is open until 24 February 2026 at UTC. More info →
    • Location: Beijing, CN

    • In-person conference organized by KCD Beijing.

    • The conference starts on the 30 December 2025.

    • Apply here
  6. 19

    days

    Kubernetes Community Days Toronto Canada 2026

    The Call For Paper is open until 17 February 2026 at UTC. More info →
    • Location: Toronto, CA

    • In-person conference organized by KCD Toronto.

    • The conference starts on the 13 May 2026.

    • Apply here
  7. 8

    days

    Kubernetes Community Days Texas 2026

    The Call For Paper is open until 6 February 2026 at UTC. More info →
    • Location: Austin, TX, USA

    • In-person conference organized by KCD Texas.

    • The conference starts on the 15 May 2026.

    • Apply here
  8. 50

    days

    Cloud Native Days Amsterdam

    The Call For Paper is open until 20 March 2026 at UTC. More info →
    • Location: Amsterdam, NL

    • In-person conference organized by Cloud Native Amsterdam.

    • The conference starts on the 22 May 2026.

    • Apply here
  9. 33

    days

    Kubernetes Community Days Panama 2026

    The Call For Paper is open until 3 March 2026 at UTC. More info →
    • Location: Panama City, PA

    • In-person conference organized by KCD Panama.

    • The conference starts on the 20 April 2026.

    • Apply here

Until next time!

— Gulcan

Subscribe to Learn Kubernetes Weekly

Trusted by 77K engineers. Delivered 168 issues and counting.

or subscribe via