en
Feedback

Don't get caught by a cheater! Telemetrio finds and tags such channels 👉 If you want to see the tag, subscribe

Kube Architect

Kube Architect

Open in Telegram

News and links on architecting and developing apps on Kubernetes curated by the @Learnk8s team

Show more
8 891
Subscribers
-424 hours
-247 days
-7130 days
Posts Archive
AX is a declarative orchestrator that runs large numbers of autonomous agent tasks in cluster sandboxes, connects their workspaces and models, and manages them with a Kubernetes-like CLI. More: https://ku.bz/zJrsrK5Qk

KitOps packages models, datasets, code, and settings into versioned OCI ModelKits, so AI and ML teams can move the exact same
KitOps packages models, datasets, code, and settings into versioned OCI ModelKits, so AI and ML teams can move the exact same bundle through development, review, and production. More: https://ku.bz/GCsMjs396

Repost from N/a
Dave Masselink, Software Engineer and Founder at Compute Gardener, compares different approaches to carbon-aware computing in the cloud-native space. Dave discusses how KEDA uses carbon intensity signals for autoscaling decisions, while Kube Green takes a different approach by snapshotting and sleeping workloads during off-hours (like weekends). He emphasizes that these tools address different aspects of the sustainability challenge - scheduling, scaling, and resource hibernation - and can work together rather than competing. Watch the full episode: https://ku.bz/zk2xM1lfW

Kube-Argus is a single-binary Kubernetes dashboard for live cluster health, pod logs, metrics, cost checks, and time-limited
Kube-Argus is a single-binary Kubernetes dashboard for live cluster health, pod logs, metrics, cost checks, and time-limited exec access without a separate database. More: https://ku.bz/22kpwpp49

Repost from N/a
Seven provisioning models create seven sets of operating responsibilities. Molly Sheets explains why one central Kubernetes team can become expensive when it owns every tenancy model across a large enterprise. Business-unit ownership with central guidance is one way to match platform responsibilities to different business goals. She discusses this staffing trade-off with Daniele Polencic. Plan the responsibilities and staffing costs with the tenancy strategy.

MariaDB Operator manages MariaDB servers and Galera clusters through Kubernetes resources, including backups, restores, upgrades, users, and databases. More: https://ku.bz/5lb6hZVGy

Repost from LearnKube news
New instructor-led course: Building GPU Platforms on Kubernetes. NVIDIA announced NVIDIA Kubernetes Engine (NVKE) at KubeCon
New instructor-led course: Building GPU Platforms on Kubernetes. NVIDIA announced NVIDIA Kubernetes Engine (NVKE) at KubeCon this week. Our new course connects NVKE to the GPU allocation, sharing, isolation, and efficiency decisions that platform teams face. You will learn how to: - Trace GPU access through NVIDIA drivers, container runtimes, and Kubernetes device allocation. - Compare time-slicing, MPS, and MIG, including their memory, compute, and fault-isolation guarantees. - Connect DCGM metrics to application throughput and identify wasted GPU capacity. - Explore NVKE workload clusters and node pools through its local evaluation environment. The four-day outline covers Kubernetes foundations, architecture, networking, and scheduling over the first three days. Day four focuses on NVIDIA GPUs and NVKE. Explore the curriculum and discuss training for your team: https://learnkube.com/gpu-kubernetes-training

This tutorial builds a two-cluster Crossplane lab, sends resources to local and remote clusters, then breaks the setup so you
This tutorial builds a two-cluster Crossplane lab, sends resources to local and remote clusters, then breaks the setup so you can learn how to read status and failure signals. More: https://ku.bz/MtkQQWrtl

This tutorial shows you how to build a Kubernetes-native LLM gateway with agentgateway, including model routes, rate limits,
This tutorial shows you how to build a Kubernetes-native LLM gateway with agentgateway, including model routes, rate limits, prompt guards, and OpenTelemetry monitoring. More: https://ku.bz/xz_rLrD2L

Repost from LearnKube news
Learn Kubernetes Weekly 209 is here. Five reads on reliable Kubernetes operations: 🛑 SIGTERM Doesn’t Mean “Stop” 💡 K8s Turn
Learn Kubernetes Weekly 209 is here. Five reads on reliable Kubernetes operations: 🛑 SIGTERM Doesn’t Mean “Stop” 💡 K8s Turn Off The Light 🔎 Your CI Pipeline Should Reject Any Service Without an Observability Contract 🌐 GKE IP exhaustion fixed: the Class E migration guide 🔗 Istio- MultiCluster MultiPrimary MultiNetwork Service Mesh on AWS EKS Read the issue: https://kube.today/issues/209 This newsletter is brought to you by LearnKube — build GPU platforms on Kubernetes. Learn GPU sharing, isolation, and efficiency through private, hands-on training for your team.

This case study shows how a team replaced CPU-based HPA scaling with KEDA, using request rate, connection pressure, and queue depth to scale event-driven workloads before they overload. More: https://ku.bz/93WBc_br6

Repost from N/a
Stefan, Principal Consultant @ ControlPlane, reflects on FluxCD v1's limits and the rewrite that changed everything. As GitOps adoption surged, v1 hit scaling ceilings. The team responded with a full rewrite — Flux v2 — built to scale, add features, and keep GitOps simple.

Repost from N/a
Phil Estes, Principal Engineer @ AWS, shares the inside story of how Containerd emerged from the early tensions between Docker and Kubernetes communities in 2014-2017. He explains how conflicts around container orchestration architecture led to the creation of Containerd as a neutral solution - a simple core runtime that could serve both Docker's commercial interests and Kubernetes' need for a runtime without built-in orchestration concepts. Phil provides context on the friction between competing visions for container orchestration and how the community navigated these challenges to create a shared foundation that ultimately benefited everyone in the ecosystem.

Repost from LearnKube news
HolmesGPT is a CNCF sandbox AI agent that investigates production incidents and finds root causes, reading from Prometheus, Grafana, Datadog and Kubernetes, and it can also run in the background 24/7. More: https://ku.bz/vDfDWlPBw

Repost from Kube Builders
Spin Operator enables Wasm-based Spin apps to run on Kubernetes using custom CRDs and a controller built with Kubebuilder. It
Spin Operator enables Wasm-based Spin apps to run on Kubernetes using custom CRDs and a controller built with Kubebuilder. It uses containerd shims for sandboxed runtime and supports local and remote clusters. More: https://ku.bz/vT-jJVvXt

Repost from N/a
Alexis Richardson, CEO @ ConfigHub, and former CEO of Weaveworks, explains how Flux’s declarative, pull-based model turned into a compliance advantage for banks and telcos. By introducing an idempotency barrier, a clear line between development and production, teams could trust that clusters pulled only from a single source of truth instead of risky ad-hoc pushes. Alexis also reflects on what was missing at the start: approvals and workflows. Enterprises later added these through pre and post-deployment hooks.

Repost from N/a
Hannah Foxwell, Co-founder at BIMP, is one of the featured lightning-talk speakers at LearnKube Day Salt Lake City 2026. Join the community for YAML Games and a reception with live DJ Amin Astaneh. Learn more: https://learnkube.com/learnkube-day-salt-lake-city-2026

Repost from N/a
LearnKube Day brings Kubernetes practitioners together for five-minute lightning talks and practical lessons from the community. Andy Suderman, CTO at Fairwinds, is one of the featured lightning-talk speakers at LearnKube Day Salt Lake City 2026. Join the community for YAML Games and a reception with live DJ Amin Astaneh. Learn more: https://learnkube.com/learnkube-day-salt-lake-city-2026

Repost from N/a
LearnKube Day brings Kubernetes practitioners together for five-minute lightning talks and practical lessons from the community. Nabarun Pal, Principal Software Engineer at Broadcom, is one of the featured lightning-talk speakers at LearnKube Day Salt Lake City 2026. Join the community for YAML Games and a reception with live DJ Amin Astaneh. Learn more: https://learnkube.com/learnkube-day-salt-lake-city-2026

Repost from LearnKube news
This week on Learn Kubernetes Weekly 208: ⚖️ Go, Node.js, C#, Java, and Python on Kubernetes 🗃️ Why Kubernetes depends on et
This week on Learn Kubernetes Weekly 208: ⚖️ Go, Node.js, C#, Java, and Python on Kubernetes 🗃️ Why Kubernetes depends on etcd and Raft 🧠 JVM memory: high usage does not always mean a leak ⏳ Dagster bottlenecks that more nodes cannot fix 🔀 Apache Stateful Functions on Flink 2.x and Java 21 Read it: https://kube.today/issues/208 ⭐️ This newsletter is brought to you by LearnKube — understand how Kubernetes works, and what to do when it breaks. Live training with 60% hands-on labs: https://ku.bz/hypSbyc-V