Looking for a way to simplify deploying LLMs on Kubernetes? This project provides everything you might need.

llmaz is an inference platform that integrates various Open Source projects for running LLMs. It supports:

- Different inference backends: vLLM, llama.cpp, Ollama, Text-Generation-Inference, SGLang, and TensorRT-LLM.
- Different model providers: HuggingFace, ModelScope, and ObjectStores.
- Chatbot interface based on Open WebUI.
- Heterogeneous devices.
- Distributed inference via multi-host and homogeneous xPyD support with LeaderWorkerSet.
- Envoy AI Gateway for token-based rate limiting, model routing, and more.
- Horizontal Pod scaling (HPA) and node autoscaling (Karpenter).

▶️ GitHub repo

Language: Go | License: Apache 2.0 | 278 ⭐️

#tools #genai

If you’re not overwhelmed yet with your work after a holiday break… or if you’re just into having some educational and practical fun with Kubernetes, don’t miss this project!

K8sQuest is a new gamified training platform for K8s, where you need to troubleshoot and fix various issues using a GUI terminal featuring arcade game styling. Importantly, it can be self-hosted locally. The project comes with:

- 50 challenges, 5 categories, 3 difficulty levels;
- different K8s topics covered, including basics, scaling, networking, storage, and security;
- progressive hints and step-by-step guides, points for completed challenges, and progress auto-saving.

▶️ GitHub repo
💬 Reddit announcement

Language: Shell, Python | License: Apache 2.0 | 326 ⭐️

#tools #fun #career

Kubernetes Dashboard is getting archived

Yesterday, the Kubernetes Dashboard maintainers announced that the “project will be archived and sunset in the coming days/weeks.” It has been developed in the Kubernetes SIG UI but lacked active contributors and maintainers for a while.

The authors recommend Headlamp as an alternative to Kubernetes Dashboard, since it became a Kubernetes SIG UI project last year.

#news #gui

Percona Everest becomes OpenEverest

Percona Everest is a Cloud Native database platform that helps operating PostgreSQL, MongoDB, and MySQL databases in Kubernetes environments. Originating as a vendor-owned solution, it has now evolved into an independent Open Source project called OpenEverest. Percona has also formed a subsidiary, Solanica, that is fully focused on developing OpenEverest. The authors plan to donate OpenEverest to CNCF soon.

#news #databases

Feel a need to validate your Dockerfiles against best practices? Consider trying this new tool.

Dockadvisor is a fast linter for Dockerfiles that helps you keep them optimal and consistent. Here’s what it offers:

- applying 60+ rules that cover standard instructions (FROM, RUN, ENV, etc.) and multi-stage builds;
- performs security checks, such as specifying secrets in variables;
- scores the quality of your Dockerfile on the scale from 0 to 100;
- can be used as a web interface, Go library, or WebAssembly module (i.e. executed in the browser).

▶️ GitHub repo

Language: Go | License: Apache 2.0 | 70 ⭐️

#tools

KubeAcademy by VMware is retired

KubeAcademy was a free educational resource for learning Kubernetes online, curated by the VMware experts. In 2024, it offered dozens of free courses. In late 2023, Broadcom finalised its acquisition of VMware, which is what most likely affected KubeAcademy's existence: “As of January 1st, 2026, KubeAcademy has been officially retired and the site will no longer be maintained or supported.”

The educational content produced by KubeAcademy is being published on GitHub.

#news #career

As many are aware, MinIO has recently gone to "maintenance mode.” If you’ve been looking for a Kubernetes-friendly alternative, check out this new project.

Garage, an S3-compatible distributed object storage, just got an unofficial Kubernetes operator. Still being in its alpha, it simplifies deploying and maintaining Garage clusters with the following features implemented:

- Deploying StatefulSets with proper configuration, storage, and networking;
- Bucket creation with quotas;
- S3 key management with automatic credential generation;
- Multi-cluster federation by connecting Garage clusters across K8s instances.

▶️ GitHub repo
💬 Reddit announcement

Language: Go | License: Apache 2.0 | 44 ⭐️

#tools #storage #news

New Kubernetes working group: Checkpoint/Restore WG

The newly announced Kubernetes WG will focus on the Checkpoint/Restore in Userspace (CRIU) ecosystem for K8s. It includes the CRIU tool as well as checkpointctl, criu-coordinator, and checkpoint-restore-operator.

You can find the charter, which defines the scope and governance of this working group, and other information, including its public meetings, on GitHub.

#news

Happy to share our first-in-2026 digest of the prominent software updates in the Cloud Native ecosystem!

1. Apache CloudStack, an Open Source IaaS solution, updated its CloudStack Kubernetes Provider, a cloud controller manager to facilitate K8s deployments, to v1.2.0. It added support for network ACLs for LB on VPC networks, a configurable source CIDR list, and ARM64 support for Docker images.

2. Harvester, a hyperconverged infrastructure solution built on Kubernetes, released v1.7.0 with experimental automatic VM workload rebalancing, support for MIG-backed vGPU devices, multipath device recognition and management, NIC hot plugging and hot unplugging, Open Virtual Format (OVF), pausable node upgrades, transparent hugepages configuration, VM VLAN trunking, and volume snapshots in guest clusters.

3. Jaeger (a CNCF Graduated project) reached v2.14.0, which added a dark theme to the UI, removed legacy v1 components (query, collector, ingester), and added a bunch of experimental features, most of which are related to ClickHouse and the FindTraces implementation for this storage.

4. Keycloak, an identity and access management solution (a CNCF Incubating project), released 26.5.0 (with 26.5.1, which followed shortly). It introduced several new features in preview, such as Workflows to automate administrative tasks, JWT Authorization Grants, exporting logs and metrics to OpenTelemetry collectors, and authenticating clients with Kubernetes service account tokens. Other release highlights are support for Caddy as a reverse proxy provider for client certificate authentication, organisation invitation management, and a guide on integrating Keycloak with MCP servers.

5. Envoy (a CNCF Graduated project) was updated to v1.37.0, which brought many new features. Some of them include global module loading and streaming HTTP callouts to HTTP filters in dynamic modules, container-aware CPU detection, new MCP filter and router for agentic network, new stats-based access logger, production-ready Proto API Scrubber filter, cluster-level retry policies, hash policies, and request mirroring, and many more.

6. Kubebuilder, an SDK for building Kubernetes APIs using CRDs, has seen its v4.11.0 release. The helm/v1-alpha projects are now automatically migrated to helm/v2-alpha, which got numerous improvements, including nodeSelector, affinity, and tolerations support, standard Helm labels for generated resources, and custom resources added to templates/extras. Newly generated projects also got their AGENTS.md files.

#news #releases

Another bunch of interesting Kubernetes-related articles recently spotted online:

1. "It works on my cluster: a tale of two troubleshooters" by Liam Mackie, Octopus Deploy.

Kubernetes has a gift for making simple problems look complicated, and complicated problems look simple. When something breaks, you often see symptoms completely unrelated to the real cause of the problem. This leads to a problem I like to call “blaming the network team”, where problems end up being diagnosed by the wrong engineers for a given issue. [..] I’ve personally experienced this dichotomy during my time as an engineer, working on both software and infrastructure, so I’m going to tell a story from two perspectives.

2. "A Brief Deep-Dive into Attacking and Defending Kubernetes" by Alexis Obeng.

My main motivation for writing this was to better understand for myself how Kubernetes works and its attack surface. I was also inspired from talking to people in the field and realizing just how prominent Kubernetes is in corporate environments. Although I did not cover every single attack vector here, I still cover a large amount of topics in the hope that this will prove useful to others seeking to understand Kubernetes’ attack surface.

3. "Exploring Cloud Native projects in CNCF Sandbox. Part 5: 13 arrivals of January 2025" by Dmitry Shurupov, Palark.

Learn about the following new CNCF projects: Podman Container Tools and Podman Desktop, bootc, composefs, k0s, KubeFleet, SpinKube, container2wasm, Runme Notebooks for DevOps, SlimFaas, Tokenetes, CloudNativePG, and Drasi.

4. "The Real State of Helm Chart Reliability: Hidden Risks in 100+ Open‑Source Charts" by Prequel.

Prequel's reliability research team audited 105 popular Kubernetes Helm charts to reveal missing reliability safeguards. The average score was ~3.98/10. 48% (50 charts) rated "High Risk" (score ≤3/10). Only 17% (18 charts) were rated "Reliable" (≥7/10).

5. "Reclaiming underutilized GPUs in Kubernetes using scheduler plugins" by Lalit Somavarapha, Gernot Seidler, Srujana Reddy Attunuri (HPE).

The default Kubernetes preemption mechanism (DefaultPreemption) can evict lower-priority pods to make room for higher-priority ones. But it only considers priority — not actual utilization. Pods are treated equivalently from a preemption perspective when they share the same priority, regardless of their current utilization. We evaluated several existing approaches.

6. "How We Built Our Deployment Pipeline: GitOps, ArgoCD, and Kubernetes at Dodo Payments" by Ayush Agarwal, Dodo Payments.

The investment in GitOps pays off at a certain scale. Below that scale, simpler solutions work fine. For us, running a payment platform with strict requirements around security, auditability, and reliability — GitOps isn’t optional. It’s infrastructure.

#articles

Kubernetes-based alternatives to Heroku are real. Here’s one of them.

Canine positions itself as a “developer-friendly PaaS for your Kubernetes”. It’s focused on small development teams and simplifies using Kubernetes for them by providing:

- container builds performed via Docker BuildKit or Buildpacks;
- automatic deployment to GitHub and GitLab;
- web UI to deploy, scale, and manage (e.g., configure resource constraints) apps running in Kubernetes;
- integration with existing K8s tools, such as Helm, cert-manager, and Telepresence;
- single sign-on via SAML, OIDC, and LDAP.

▶️ GitHub repo

Language: Ruby | License: Apache 2.0 | 2716 ⭐️

#tools #gui

ClickHouse just got the official Kubernetes operator

Less than 5 hours ago, the official ClickHouse Operator got its first public release, v0.0.1. It allows you to create and manage ClickHouse clusters and features ClickHouse Keeper integration, storage provisioning, TLS/SSL support, and Prometheus metrics integration.

The operator is written in Go, is Open Source (Apache 2.0) and available on GitHub.

#news #releases #databases

vCluster introduced vind, marketed as a better kind

vCluster Labs (previously known as Loft Labs) released a new tool called vind (vCluster in Docker). It is built on top of vCluster and allows you to run Kubernetes clusters directly as Docker containers, similarly to what kind (Kubernetes IN Docker) offers. However, it comes with the following extra features:

- pausing the clusters when they're not in use and resuming them;
- automatic LoadBalancer support;
- image caching (pull-through cache via local Docker daemon);
- support for connecting external nodes, which can be real cloud instances;
- support for choosing CNI and CSI plugins;
- built-in vCluster Platform UI.

You can find more details about vind on GitHub and in yesterday’s video presentation on LinkedIn.

#news #tools

Optimising resources in Kubernetes is something we all want to do at some point. This new project aims to assist in that.

CruiseKube, dubbed as “Autopilot for Kubernetes”, is a controller that watches your K8s workloads and adjusts the resources accordingly. Here’s what it does:

- Continuously evaluates current CPU/memory usage and updates resource requests.
- Considers CPU pressure (PSI metrics) and other Pods on the node when resizing.
- Watches OOM memory values in stats and triggers Pod eviction when needed.
- Uses Prometheus as the primary metrics source.
- Provides a web UI to see and manage your settings.

▶️ GitHub repo
💬 Reddit announcement

Language: Go | License: MIT | 48 ⭐️

#tools

Node Readiness Controller for Kubernetes

Last week, a new Kubernetes SIG project was announced. The Node Readiness Controller can be used to define additional requirements for node readiness (e.g., GPU drivers are loaded). The controller will manage node taints to prevent scheduling until the required conditions are satisfied. It supports bootstrap-only and continuous enforcement modes. Currently, the project is in its alpha.

Find more details in the project’s documentation and on GitHub.

#news #tools

Our latest selection of interesting Kubernetes-related articles recently spotted online:

1. "Kubernetes Rolling Updates for Reliable Deployments" by James Walker, Spacelift.

In this guide, we will explain the benefits of rolling updates, describe how they work, and provide detailed examples of their use. We’ll also compare how rolling updates stack up against other popular deployment strategies.

2. "Experimenting with Gateway API using kind" by Ricardo Katz, Red Hat.

This document will guide you through setting up a local experimental environment with Gateway API on kind. This setup is designed for learning and testing. It helps you understand Gateway API concepts without production complexity.

3. "Understanding the Ingress-NGINX Deprecation — Before You Migrate to the Gateway API" by Artem Lajko.

Most blog posts about the Ingress-NGINX deprecation are optimized for clicks, not for engineers who actually have to migrate production systems. You’ll find tiny demo setups, toy examples, and conclusions that fall apart the moment you apply them to an enterprise environment. That frustration is the reason this guide exists. This article is based on our real enterprise setup, built on top of the kubara framework. It documents how we approached the migration, what worked, what didn’t, and — just as important — what we decided not to migrate.

4. "Lazy-Pulling Container Images: A Deep Dive Into OCI Seekability" by Zain Malik.

This post starts with why the problem is harder than it looks at the byte level, then surveys the major approaches and what they trade off. The core of the post is a hands-on experiment: I deploy an in-cluster registry, convert images to eStargz, patch containerd with a custom snapshotter, and measure something nobody benchmarks properly. Not just pull time, but readiness, the moment a container can actually serve its first request.

5. "Kernel Archaeology: Why 36 CPUs Crash Cilium But 32 Don’t" by Pierre Magne, Qonto.

[..] The deployment looked successful. But then, over several weeks, we noticed sporadic crashes — roughly one Cilium agent per week, completely unrecoverable without restarting the entire node. No clear pattern, no obvious trigger. Rare enough to be hard to reproduce, but severe enough to block production deployment.

6. "Speeding Up FluxCD Development Without Remote Pushes: Local Git Reconciliation" by Marco Boss.

[..] I started looking for a way to develop and validate manifests locally, while still having full access to Flux features, and without resorting to brittle hacks or partial simulations. In this post, I’ll walk you through the approach I ended up with and show you how to run Flux locally in a way that actually feels usable for day-to-day development.

#articles

Kubernetes WG Serving is disbanded

Yuan Tang, on behalf of the Serving working group co-chairs, announced that the WG Serving’s goal had been accomplished and that the group is disbanded.

WG Serving was created to support the development of the AI inference stack on Kubernetes, making it "an orchestration platform of choice for inference workloads". In particular, it contributed to the design of AIBrix (a part of vLLM), while other unresolved problems were implemented by llm-d. The working group also helped with Kubernetes AI Conformance requirements.

All existing related efforts are now covered by other SIGs and working groups (including SIG Node, SIG Scheduling, and WG Device Management) or specific projects (such as Gateway API Inference Extension and Inference Perf).

#news #genai

Sharing our latest digest of the prominent software updates in the Cloud Native ecosystem!

1. Longhorn, a Cloud Native distributed storage for Kubernetes (a CNCF Incubating project), released v1.11.0, which brought its V2 Data Engine to the Technical Preview. Other updates include balance-aware algorithm disk selection for replica scheduling, active monitoring for node disk health, and support for Kubernetes RWOP (ReadWriteOncePod) and StorageClass allowedTopologies.

2. Argo CD (a CNCF Graduated project) reached v3.3.0. It introduced PreDelete hooks, automatic background refresh of OIDC tokens, support for resource names in clusterResourceWhitelist, shallow cloning for repositories, and KEDA support (pausing and resuming KEDA resources from the Argo CD UI and ScaledJob health checks).

3. Headlamp, a Kubernetes web UI developed by the Kubernetes SIG, was updated to 0.40.0 and got configurable keyboard shortcuts, HTTPRoute support for Gateway API, icon and colour configuration for clusters, saving selected namespaces per cluster, support for a8r.io service metadata in service views, and more.

4. Cilium (a CNCF Graduated project) released 1.19.0 with lots of new features. They include multi-level subdomain matches in DNS policies, support for VRRP and IGMP protocols in host firewall rules, strict encryption modes for both IPsec and WireGuard, enrolling namespaces into Ztunnel, support for GRPCRoute in GAMMA, TLS/mTLS support for Prometheus metrics, and CRD auto-installation for Multi-Cluster Services.

5. KEDA (a CNCF Graduated project) was updated to v2.19.0, introducing a new Kubernetes Resource Scaler, file-based authentication support for ClusterTriggerAuthentication, and other improvements.

6. Karpenter, a Kubernetes Node Autoscaler developed by the Kubernetes SIG, unveiled its v1.9.0, adding Gte and Lte operators for requirements, a NodePool cost metric, and consolidation pipeline logging.

7. Istio (a CNCF Graduated project) 1.29.0 was released with DNS capture and iptables reconciliation enabled by default for ambient workloads, CRL (Certificate Revocation List) support in Ztunnel, debug endpoint authorisation enabled by default, alpha support for wildcard hosts in ServiceEntry resources with DYNAMIC_DNS resolution, HTTP compression for Envoy metrics, pilot resource filtering capabilities, and many other changes.

#news #releases