Facilitating the Spread of Knowledge and Innovation in Professional Software Development

Write for InfoQ


Choose your language

InfoQ Homepage News AI, Orchestration, Native Network and K8sGPT: KubeCon EU Highlights New CNCF Sandbox Projects

AI, Orchestration, Native Network and K8sGPT: KubeCon EU Highlights New CNCF Sandbox Projects

This item in japanese

As highlighted at the recent KubeCon EU 2024 conference, several new projects joined the CNCF sandbox in December across a range of categories: kube-burner in CI/CD, Kuasar in the container runtime, K8sgpt in the observability, KRKN in chaos engineering, easegress in API Gateway, spider pool in Cloud-native network and KubeStellar in Scheduling and Orchestration.

Jorge Castro, open source community manager and developer relations at the CNCF, stated at the event that "the CNCF sandbox is the shiny new toys department of the foundation, where you can find the next thing to try out or to contribute to". InfoQ explored several of these new CNCF projects.

Kube Burner is a Kubernetes performance and scale test orchestration toolset. It provides multi-faceted functionality, including:

  • Create, delete and patch Kubernetes resources at scale
  • Prometheus metric collection and indexing
  • Measurements and alerting

Kube-burner is written on top of the official k8s client library, client-go. At the time of writing, its repository counted 422 stars and 39 contributors. Its first commit was submitted in August 2020.

KubeStellar: Defines itself as the "post-office" for Kubernetes resources. It will take a package and deliver it to the right recipient, whether in public, private or edge clouds. Asked about the mission, Donny Rose, a contributor to the project, stated:

We make it as easy to deploy and configure multiple clusters as it is for one using all the K8s tools you are already familiar with


The project promises to treat multiple Kubernetes clusters as one, enabling engineers to:

  • Centrally apply Kubernetes resources for selective deployment across multiple clusters
  • Use standard Kubernetes native deployment tools (kubectl, Helm, Kustomize, ArgoCD, Flux); no resource bundling is required
  • Discover dynamically created objects created on remote clusters
  • Make disconnected cluster operation possible
  • Scale with 1:many and many:1 scenarios
  • Remain compatible with cloud-native solutions

KubeStellar’s repository counted 205 stars and 35 contributors. Its first commit was submitted in November 2022.

Easegress is a cloud-native proxy designed for highly available and highly performant traffic orchestration. The system is built with observability, extensibility and integration in mind.

The Easegress repository counts 5.7 K stars and 63 contributors. The first commit was submitted in March 2017.

Kuasar is an efficient container runtime that "provides cloud-native, all-scenario container solutions by supporting multiple sandbox techniques". The project supports MicroVM, WASM, App Kernel and runC as sandboxed techniques.


The repository has 1.1 K stars and 22 contributors. The first commit was submitted in April 2023.

Spiderpool is an underlay and Remote Direct Memory Access (RDMA) network solution designed for Kubernetes environments. The tool can be used exclusively or in shared mode, and it is useful for both IPv4 and IPv6 permitting different combinations: IPv4-only, IPv6-only or both. The implementation is based on RDMA over Converged Ethernet(RoCE) and InfiniBand technologies.

Weizhou Lan: Spiderpool is useful for AI workloads as it will help reduce the training time by bringing resources together in an efficient manner through RDMA

The repository has 436 stars and 35 contributors. The first commit was submitted in March 2022.

K8sGPT: Defines itself as a tool for scanning Kubernetes clusters, and diagnosing and triaging issues in simple English. K8sGPT has SRE experience codified into its analyzers, which helps to pull out the most relevant information that is then enriched using generative AI. All of the interactions can be conducted directly in the terminal.

When asked about the particularities of the model, Thomas Schuetz, a contributor to the project, said: "We are just anonymizing the data and sending it to the AI. There was no fine-tuning done".

Currently, K8sGPTIt supports models like OpenAI, Azure, Cohere, Amazon Bedrock, Google Gemini and local models. It runs on Windows, Mac and Linux machines and can be installed via brew, RPM, DEB or APK. The tool can also be installed in a Kubernetes Cluster.

The project was among the top 10 most contributed projects from CNCF. Currently, the repo has 4.7K stars and 69 contributors. The first commit was in March 2023.

KRKN defines itself as a "chaos and resiliency testing tool for Kubernetes with a focus on improving performance under failure conditions". The tool injects failures into Kubernetes clusters to check if it is resilient to turbulent conditions. The recommended way to get started is by using the Chaos Recommender, a utility tool that analyses telemetry data (CPU, network, memory) and recommends running chaos scenarios afterwards. The tool currently supports scenarios from pod to network chaos or managed cluster scenarios.


KRKN can be integrated into a CI pipeline and supports both private and public clouds. The repo currently has 234 stars and 38 contributors. The first commit was submitted in April 2020.

The CNCF tooling ecosystem continues to grow and now contains 184 projects. The sandbox category includes 114. Engineers are encouraged to experiment and to provide feedback or contribute.

About the Author

Rate this Article