"Our main shutdown mode is a hard power cut — when the driver switches off the engine."
Jordan Karapanagiotis on running Kubernetes on a tractor
📺: https://ku.bz/3yDmJ_xSC
b4n is a k9s-inspired terminal UI for Kubernetes written in Rust, built on kube-rs and ratatui with full CRUD, log streaming, shell access, and port forwarding
➜ https://ku.bz/d-5zDDPY_
"People are very effective at standing up AI workloads. Then they look at the cost and say, wow."
Andrew Hillier on why GPU cost optimization is still DIY
📕: https://ku.bz/KL4jRvsL4
https://miro.medium.com/v2/1*mBcpZTH1ssJtLbc3h-T1hw.png
This case study shows how Wayground handles 7,000 frontend deployments per month across 50 engineers using GitOps and ArgoCD with zero deployment queues or approval processes
➤ https://ku.bz/ykdMXw-_S
This Cluster API provider lets you create and manage Kubernetes clusters running on Incus virtual machines using declarative Kubernetes-style configurations
➤ https://ku.bz/ds2ynCZPP
https://assets.learnk8s.io/linkedin-174.png
This week on the Learn Kubernetes Weekly:
🤖 75% Faster Debugging with AI
🔥 60% Less Pods, Doubled Traffic Capacity
📈 Scaling Django to 1M Users
⚠️ Kubernetes Bad Practices
🥷 PKI Credential Abuse
⭐️ LearnKube
Read it now: https://kube.today/issues/174
https://kube.careers/image-gen/digest?companies=Anthropic&companies=Parloa&companies=Anduril+Industries¤cy=USD&salaryFrom=171400&salaryTo=87510466&subtitle=This+week&title=6+Kubernetes+jobs+that+offer+relocation
HAProxy Data Plane API is a sidecar process written in Go that runs alongside HAProxy and provides REST API endpoints for managing HAProxy configuration, requiring HAProxy version 1.9.0 or higher
➜ https://ku.bz/wyGW4WtdG
Cluster API provides a powerful abstraction layer for managing Kubernetes clusters across different cloud providers, explains Zain Malik. It enables multi-cloud capabilities through standardized abstractions
Watch the full episode: https://ku.bz/5PLksqVlk
https://miro.medium.com/v2/1*7iI7YlA9ERrVzAxJ1dkpQg.png
This tutorial teaches how to configure GPU time slicing on AWS EKS using Karpenter to provision Bottlerocket nodes
This allows multiple pods to share a single physical GPU
➤ https://ku.bz/-hsSTdMK2
Actions Runner Controller (ARC) is a Kubernetes controller for GitHub Actions self-hosted runners
With ARC, you can:
- Deploy self-hosted runners on Kubernetes clusters with a simple set of commands
- Auto scale runners based on demand
➜ https://ku.bz/ZjFXy9Czk
🗣️ Fernando from SadServers on how he cut his Kubernetes bill from $1,000/month on GKE to $30/month on Hetzner with Edka — a 500% cost reduction for the same capacity
https://ku.bz/6nSDbz9m4
🌟 LearnKube
🎙 🎙Bart
This CSI driver enables mounting cloud storage providers supported by rclone as persistent volumes in Kubernetes clusters
➤ https://ku.bz/czJ7TMvBX
https://miro.medium.com/v2/0*XIeJGF7iQrF91hUn.png
This case study describes rebuilding a fragile Kubernetes infrastructure into a production-grade platform for GPU-based ML workloads, improving deployment frequency from weekly to 10+ times daily
➜ https://ku.bz/lFNBx-8yT
"A lot of organizations don't have that choice. You basically have to take what your security team is telling you."
Mauro Morales on picking an OS for Kubernetes
📺: https://ku.bz/8cpgjFfjn
Argo Workflows is fire-and-forget. Once the job's done, pods terminate—no streaming
That's why Sriharsha Yayi and Derek Wang built NumaFlow
📺: https://ku.bz/dbFK-9Yxw
📣: https://ku.bz/PQ-hn3ZCm
This code lets containerd pull and run images directly from the Nix store instead of traditional OCI layers by giving containerd native awareness of Nix packages and a CRI image service for Kubernetes
➤ https://ku.bz/77WMMFTdq
https://github.com/skyhook-io/radar/raw/main/docs/screenshot.png
Radar provides Kubernetes cluster visibility through topology graphs, event timelines, and service traffic visualization running as a single binary that connects directly to the Kubernetes API without cluster-side installation
➜ https://ku.bz/MhXx4kyjB
"We are very excited to announce the launch of Hadron Linux — a Linux distribution engineered from scratch by the Kairos team."
Ettore Di Giacinto on the new minimal OS for edge Kubernetes
📺: https://ku.bz/wMhKpZ5bQ
📣: https://ku.bz/_9RmXnjDJ
https://miro.medium.com/v2/1*No0EJIUEz3dmwqN-tm607w.png
This case study walks through how the team moved Argo CD from an in-cluster Redis to AWS ElastiCache Serverless, which almost wiped out their Redis costs and removed all the maintenance work
➤ https://ku.bz/TmbtK2v0K
https://miro.medium.com/v2/1*WAqMiw6h23D2Snn5ykGyLg.png
This tutorial teaches how to build a home Kubernetes cluster using three Raspberry Pi 4 devices for hands-on learning and experimentation
➜ https://ku.bz/M2ctfBBpC
Korrel8r is a rule-based correlation engine that automatically discovers and graphs relationships between Kubernetes cluster resources and observability signals across Prometheus, Loki, Alertmanager, and the Kubernetes API
➤ https://ku.bz/tdFklzjTN
H8s is a home infrastructure project combining Kubernetes with Talos OS security, running on 2 N100 mini PCs with GitOps deployment via ArgoCD
➤ https://ku.bz/2VMpLQ5Tp
pwru is an eBPF-based tool for tracing network packets in the Linux kernel with advanced filtering capabilities
It allows fine-grained introspection of kernel state to facilitate debugging network connectivity issues
➜ https://ku.bz/Q3X1ngZGC
https://assets.learnk8s.io/linkedin-173.png
This week on the Learn Kubernetes Weekly:
🔥 Egress with Squid Proxy
💪 OS Migration Reduced Infra 30%
⚡ Auto & Load-based Scaling
🎯 Smart Scheduling: Cost Optimization
🤖 Claude Code Pilots K8s
⭐️ Hadron
Read it now: https://kube.today/issues/173
https://kube.careers/image-gen/digest?companies=Anthropic&companies=Parloa&companies=Anduril+Industries¤cy=USD&salaryFrom=171400&salaryTo=87242145&subtitle=This+week&title=6+Kubernetes+jobs+that+offer+relocation
This article shows how to migrate a Factorio game server from a basic VPS to Kubernetes with Longhorn storage, addressing lag issues and improving reliability
➤ https://ku.bz/GN8XQThYw
This article teaches how to make Kubernetes workloads survive AWS Spot Instance interruptions using NTH or Karpenter, PDBs, stateless design, and AWS FIS game days
➜ https://ku.bz/RZ4TYSl95
https://res.cloudinary.com/learnk8s/image/upload/v1772544380/gpu-sharing-problems-2026/slide-1.png
You want to share GPUs: one team runs inference, another trains models, and both need the same expensive cards.
The problem is that GPUs don't behave like CPU and RAM under contention.
(I will cover this on Thursday: ku.bz/multitenant26 )
🧵
"We assumed that implicit budget happens in Karpenter. It's not the case."
Adhi Sutandi on Karpenter's default 10% disruption budget catching his team off guard
📺: https://ku.bz/XyVfsSQPr
