[Sandbox] K8sGPT

[AlexsJones]

Application contact emails

alexsimonjones@gmail.com, thomas.schuetz@t-sc.eu

Project Summary

Kubernetes cluster analysis augmented with Artificial Intelligence

Project Description

The popularity of Kubernetes has skyrocketed, but it has also led to a significant amount of complex knowledge regarding the management of cluster workloads. As new innovations emerge, it becomes increasingly challenging to manage workloads and identify potential issues.

K8sGPT employs codified SRE techniques, utilizing Artificial Intelligence (either hosted or bring-your-own) to simplify the description of complex problems and provide easy-to-implement solutions. This tool is accessible through CLI or as an Operator and can be integrated with observability projects to facilitate continuous monitoring and straightforward triage.

The goal of K8sGPT is to act as a virtual engineer, reducing the number of personnel required on your team and eliminating two of the most significant obstacles to cloud-native adoption: cost and skill.

Org repo URL

https://github.com/k8sgpt-ai/community

Project repo URL

https://github.com/k8sgpt-ai/k8sgpt

Additional repos

https://docs.k8sgpt.ai/
https://github.com/k8sgpt-ai/k8sgpt-operator

Website URL

https://k8sgpt.ai

Roadmap

Backlog

Roadmap context

The K8sGPT initiative aims to simplify Kubernetes for end-users and streamline workload management. To achieve this goal, the project intends to enhance Observability integrations, resource analysis, and AI backends. The roadmap includes developing an interoperable system that reflects cloud-native principles and enables individuals and organizations, from small to large, to incorporate K8sGPT into their toolkits.

We are also mindful that business value is a key indicator of success and as such will ensure we liaise with end-user organisations to understand their use cases and any bars to adoption ( Local AI and anonymisation being features we already anticipated being in high levels of demand ).

Contributing Guide

https://github.com/k8sgpt-ai/k8sgpt/blob/main/CONTRIBUTING.md

Code of Conduct (CoC)

https://github.com/k8sgpt-ai/k8sgpt/blob/main/CODE_OF_CONDUCT.md

Adopters

https://github.com/k8sgpt-ai/community/blob/main/ADOPTERS.md

Contributing or Sponsoring Org

No response

Maintainers file

https://github.com/k8sgpt-ai/community/blob/main/MAINTAINERS.md

IP Policy

• If the project is accepted, I agree the project will follow the CNCF IP Policy

Trademark and accounts

• If the project is accepted, I agree to donate all project trademarks and accounts to the CNCF

Why CNCF?

Our team firmly believes that K8sGPT is a valuable addition to the CNCF's impressive range of projects. With artificial intelligence advancing rapidly, embracing this technology would demonstrate the CNCF's commitment to innovation and thought leadership in the field of cloud-native computing.

As a lighthouse project, K8sGPT showcases the potential for AI to be both cloud-native and community-driven and highlights the benefits of composability within the CNCF ecosystem. This project also promotes democratic access to AI and encourages the use of local models for exploring this exciting new technology.

We are confident that joining the CNCF community will help us expand our contributor base, enhance our project maturity, and attract more adopters. In return, we hope to offer valuable contributions to the community, including new end-users, cross-project solutions, and a smoother path to cloud-native computing.

Benefit to the Landscape

Already embracing Prometheus and SIG-Kubebuilder, K8sGPT looks to strengthen the bonds across projects and provide deep value whilst guiding our aspirations towards simplicity and ultimately our end-users.

There are many tools out there, which analyze Kubernetes Environments. We aim to add our SRE Knowledge as far as possible to simplify troubleshooting. Furthermore, K8sGPT is extensible so that other CRD-based configurations can troubleshoot as soon as a certain analyzer is available.

As one of the first AI-based troubleshooting tools in the Landscape, we believe in the power of this technology and also plan to pre-train our own models to make troubleshooting easier. Given the structure of the project, we support contributors putting their own ideas into practice and getting a home for their project.

By supporting a variety of AI model backends, including local models, we also want to send a strong message that AI is not only the preview of proprietary backends, encouraging a new wave of end-users to start experimenting.

Cloud Native 'Fit'

Born from the intersection of Artificial Intelligence and Kubernetes, K8sGPT is a project that looks to make Kubernetes (the world's most successful cloud-native project) accessible to even more communities. Students, scientists, researchers and engineers, all looking to uplift to cloud-native, now have a project that is guided by the principle of increasing accessibility.

With such a focus on easing the adoption of cloud-native principles we are providing access for those involved in trying to accelerate solutions for crime, poverty, climate change and other deep-seated societal issues that cloud-native can impact, this is our small way of being part of that change.

Cloud Native 'Integration'

K8sGPT primarily focuses on troubleshooting Kubernetes Environments. Furthermore, it is capable of integrating results from other tools into its analysis. The first tool, where this kind of integration has been implemented was Trivy.

Furthermore, the tool can be integrated into other workflows, e.g. to inspect the health of an application after it has been deployed. As it is a CLI and containers are available, it’s possible to integrate it into almost every CI tool. The first integration into a deployment workflow has been implemented with Keptn.

Cloud Native Overlap

N/A

Similar projects

N/A

Product or Service to Project separation

N/A

Project presentations

• KCD Colombia, Brad McMcoy ( Links will be made as available )
• TAG Observability ( Scheduling an available slot, links will be made as available )

Project champions

Bartłomiej Płotka (TAG Observability, Tech Lead)
Anton Smith ( Spectro Cloud, Head of Product )

Additional information

We are contributors who are very familiar with the CNCF and are proud to call many people within the community our colleagues and friends. As such, would not submit a project unless we saw a real nugget of potential and interest for the wider community. As something precious to us, we want to share with you and grow it for everyone.

( There is no company behind k8sgpt just a GitHub organisation for our community. We feel that whilst many would look to monetize this project, we want it to be for everyone. )

[pfarikrispy]

❤️👍

[rochaporto]

Hi @AlexsJones .

Following the 2023-06-13 TOC Meeting the project is seen as very interesting but we think it is still very early and recommend reapplying in 6 months.

Thanks!

[AlexsJones]

Point taken @rochaporto thanks!

Can you link me to where duration of the project is a requirement in the application for future reference?

I was part of the group successfully submitting open feature, without a line of code yet written.

[TheFoxAtWork]

@AlexsJones Apologies for the delayed response.

Currently, the TOC does not define explicit criteria for sandbox applications beyond it "being cloud native" (at least not in the same way we do for Incubating and Graduating projects). However, we've come to find over the application cycles that projects find a higher rate of success and engagement within Sandbox when they achieve a level of progress and growth at the time of application to sandbox. What that looks like varies with each project and is also impacted heavily by the demand or need from adopters and the broader ecosystem.

As you pointed out, some projects have been accepted without extensive codebases whereas others we really want to see more robustness in what it does and how it does it. It largely depends on the project and today we don't have an easy way to codify what that is for each and every project that applies. What the TOC has leaned towards to address this is restructuring the application for sandbox (as you see with the form that K8sGPT applied) and using windows of time to determine progress.

For instance, we saw project X apply on this date, but things were too nascent to make an informed decision, so in 6 months, perhaps the ecosystem/landscape will have shifted to be more informative, or the project may have smashed barriers and really grown in that time. In some cases, we've accepted projects that have dissolved mid-way through onboarding and we don't have strong indicators as to which projects this will happen to, therefore we try to leverage time a good indicator to see interest and gauge sustainability/viability of the project.