The Themes
1. There is no silver bullet for multicluster management.
Kubernetes multicluster management is a maturing space with much development and trialing taking place. At present, it tends to require a large amount of tooling and processes compared to other areas of the ecosystem. While there are many tools available, few if any stand out as comprehensive solutions. Many companies find it necessary to pick out a number of tools from the large selection available to meet their unique needs. Specific requirements may vary widely across organizations, further contributing to the diversity of tools currently in use. Show More
2. Cluster management often requires custom-built in-house solutions.
For both deploying and managing clusters, whether in managed or unmanaged Kubernetes, as well as for managing core services, one conclusion was clear: Custom tooling is usually a necessity. In fact, even managed Kubernetes tends to require external custom tooling. Show More
3. Common tool combinations include: Helm with Operators & GitOps Argo/Flux.
Tools for management of core services -- like DNS, metrics, logs, RBAC, namespaces, etc. -- outnumbered tools for cluster deployment on the Radar. Helm, Operators, GitOps, Kustomize and custom tools were the most widely used. Helm’s out-of-the-box charts for every product make it highly accessible, which helps explain its high number of "Adopts" in the Radar. We observed that Helm is often adopted together with GitOps, with users split between the latter’s Argo and Flux tools. We also saw Operators added to Helm and GitOps in a large number of companies.
4. Operators are a favorite tool for management of clusters and core services.
Operators were the standout tool in the Radar, with a total of 24 “Adopt” votes in the core services category. The consensus was that Operators make managing clusters simpler and minimize overhead. Show More
5. The community eagerly awaits readiness of Cluster API.
Many end users are closely watching the development of Kubernetes sub-project Cluster API, hoping it will solve some key infrastructure management issues and reduce the need for custom tooling. Cluster API is attempting to reduce complexity in the space through flexible methods and declarative, Kubernetes-style APIs for cluster creation, configuration and management. Show More
The End User companies
The Radar Team
Gabe Jackson
Cloud Platform Tech Lead at Mattermost. I have worked on various distributed systems over the years and generally enjoy the challenges in this space.
Federico Hernandez
Principal Engineer at Meltwater. Working with the internal platform offering based on Kubernetes. Program committee member and track chair "Operations" for a couple of KubeCons.
Simone Sciarrati
Engineering Team Lead at Meltwater, where I have been focusing on building an internal Kubernetes based platform to make life easier for our development teams.
Rajarajan Pudupatti SJ
Vice President of Cloud Platform Delivery at Fidelity Investments, responsible for the architecture and delivery of the next-gen application platform. Committed to bringing all the benefits of various cloud native technology projects to Fidelity enterprise teams, sparing them the need for significant effort on their side, and, hopefully, open-sourcing most of our work for the larger community to benefit from as well. Also a proud AWS container hero.
The data
The CNCF End User Community was asked to describe what their companies recommend for different solutions: Adopt, Trial, Assess or Hold. This table shows how the End User companies rated each technology.
The industries
| Software | 7 |
| Financial Services | 3 |
| Entertainment | 3 |
| Travel | 2 |
| E-commerce | 2 |
