Cloud Native Weekly: k0s Joins CNCF Sandbox

Open Source Project Recommendations

HAMi

HAMi (formerly known as k8s-vGPU-scheduler) is a CNCF Sandbox-level open-source middleware for Kubernetes. It virtualizes heterogeneous devices like GPUs/NPUs and supports memory, compute core timeslice isolation, and shared scheduling. HAMi provides a unified interface for containers, enabling fine-grained resource quotas and isolation without modifying the application code. This enhances device utilization and scheduling flexibility and has been widely adopted in public/private clouds and industries like finance, manufacturing, and telecommunications.

SkyPilot

SkyPilot is an open-source framework developed by the Berkeley Sky Computing Lab. It provides a unified YAML/CLI/Python interface that allows users to easily schedule AI training, inference, and batch jobs on Kubernetes or over 16 cloud platforms (such as AWS, GCP, Azure, etc.). It intelligently selects the cheapest regions (including automatic recovery of spot instances) and automatically cleans up idle resources, ensuring high GPU availability and significantly reducing costs.

Trainer

Trainer is a Kubernetes-native open-source project aimed at enabling large-scale distributed training and LLM fine-tuning across multiple ML frameworks (e.g., PyTorch, TensorFlow, JAX, XGBoost). It offers a unified Python SDK, CRDs (like TrainJob, TrainingRuntime), and controllers. Trainer integrates with ecosystems like Hugging Face, DeepSpeed, and Megatron-LM for extensibility and uses CPU-based loaders to reduce GPU initialization overhead. It supports auto-scaling, resource optimization, migration, and multi-cloud deployment for training workloads.

Envoy AI Gateway

Envoy AI Gateway is an open-source, Kubernetes-native AI gateway built on Envoy Gateway and Envoy Proxy. It provides a unified, secure, and scalable entry point for application traffic to connect with various large language models (LLMs) and generative AI services. By standardizing APIs from different providers (currently supporting OpenAI and AWS Bedrock), it enables intelligent routing, failover, token-based rate limiting, access control policies, upstream authentication (API Key, AWS Signature, OIDC, etc.), and deep integration with Kubernetes via CRDs and Helm.

Technical recommendations

Kubernetes v1.33: Updates to Container Lifecycle

This article introduces new container lifecycle features in Kubernetes v1.33, including support for zero-duration Sleep actions (enabled by default) and the Alpha phase introduction of configurable container termination signals. Sleep actions allow pause mechanisms before container start or termination, and zero values simplify configuration and image dependencies. The termination signal feature lets users explicitly specify termination signals via Pod spec, improving graceful shutdown control and cross-platform compatibility. These enhancements boost Kubernetes’ flexibility and operability in managing container lifecycles.

Milestone for Lightweight Kubernetes: k0s Officially Joins CNCF Sandbox

k0s is a lightweight, zero-dependency Kubernetes distribution deployable with a single binary. It supports use cases from developer laptops and data centers to resource-constrained edge environments. Recently, k0s officially joined the CNCF Sandbox, marking its entry into the early-stage cloud-native ecosystem. This move aims to gain broader community support, feedback, collaboration opportunities, and exposure — paving the way for easier, more accessible Kubernetes usage and transparent open-source governance.

What’s new in cloud native

Podman Desktop 1.19 Released

Podman Desktop 1.19 was recently released, updating the built-in Podman executable to v5.5 and enhancing stability and features. The update also includes improvements to several extensions: Bootc 1.9 now supports creating VMs from bootable containers, AI Lab 1.7.2 adds llama-stack and model context protocols, a new Minc extension supports MicroShift container deployment, and extensions for RHEL Lightspeed and RHEL VMs have been introduced. Additionally, Podman Desktop is now directly installable on RHEL 10.

GitLab Releases Self-Hosted GitLab 18 with AI Code Assistance

GitLab officially released the self-hosted GitLab 18.0 on June 6, 2025, introducing native AI features for Premium and Ultimate users. These include context-aware Duo code suggestions, real-time IDE chat, code generation, and refactoring to enhance code review efficiency. These AI tools are integrated into popular IDEs and the merge request workflow, supporting cross-file change analysis to reduce irrelevant suggestions. The self-hosted environment also includes Repository X-Ray for intelligent analysis and enhances security and compliance with improved SAST, vulnerability tracking, fine-grained CI/CD permissions, SHA256 SAML certificate support, and deletion protection.

About KubeSphere

KubeSphere is an open source container platform built on top Kubernetes with applications at its core. It provides full-stack IT automated operation and streamlined DevOps workflows.

KubeSphere has been adopted by thousands of enterprises across the globe, such as Aqara, Sina, Benlai, China Taiping, Huaxia Bank, Sinopharm, WeBank, Geko Cloud, VNG Corporation and Radore. KubeSphere offers wizard interfaces and various enterprise-grade features for operation and maintenance, including Kubernetes resource management, DevOps (CI/CD), application lifecycle management, service mesh, multi-tenant management, monitoring, logging, alerting, notification, storage and network management, and GPU support. With KubeSphere, enterprises are able to quickly establish a strong and feature-rich container platform.

To stay updated, visit our official website or follow us on Twitter.