L

Why Mercedes-Benz uses 900 Kubernetes clusters PCWorld

The German automaker, who runs large Kubernetes clusters to support project teams around the world, tells him that managing Kubernetes isn’t that difficult. The technical team at German automaker Mercedes-Benz has built 900 Kubernetes clusters over the past seven years to support hundreds of autonomous developer teams, providing the company with a modern, scalable infrastructure platform and easy-to-manage delivery. Automakers started using Kubernetes to deploy applications in 2015, a year after Google offered an open source container orchestration system. Since then, Mercedes-Benz Tech Innovation, the automaker’s technology subsidiary, has developed in-house expertise to support hundreds of application teams linked to its divisions according to their technical requirements. Mercedes-Benz Tech at KubeCon Europe 2022. Jens Erat, DevOps Engineer at Innovation, said, “We knew that a single shared Kubernetes cluster didn’t meet our requirements, vendor deployments didn’t meet our requirements, and we had the right engineers.” It was held from May 16th to 20th. “Our 100% FOSS (Free Open Source Software) platform was designed and developed by the same devops team, with no licensing issues or support requests,” he added. Currently, Mercedes-Benz runs on 900 on-premises Kubernetes clusters in four global data centers using OpenStack. All have been running Kubernetes version 1.23 since late 2021. According to a 2019 Cloud Native Computing Foundation (CNCF) survey, automakers’ Kubernetes domains are smaller than cloud providers, but less than 10% of enterprises use 50 or more. cluster. This domain is also nearly five times larger than the Kubernetes environment at CERN, another backbone of KubeCon Europe, which was running 210 clusters at the time of writing. What is the limit on the number of clusters? “We are working hard to keep this environment manageable,” said Peter Müller, senior expert at Mercedes-Benz Tech Innovation. “Everything is automated, so if you manage 500 clusters or 1,000 clusters, the surrounding systems work well. It works: to add 500 more clusters, you only need to add one more engineer. “What was able to solve this management problem was mainly the declarative creation, configuration and Thanks to OpenStack’s Cluster API, a Kubernetes project that makes management possible. However, as always with technology, this solution is far from perfect. “The number of clusters is not an issue. The problem is with some peripheral systems and sometimes OpenStack,” said Müller. “But Kubernetes works very well and is extensible.” Culture Change Mercedes-Benz’s hundreds of application teams can now request their own Kubernetes clusters through an automated process using an in-house toolset built and maintained by the Peter Müller team within Mercedes-Benz Tech Innovation. . Teams typically request pre-provisioned production clusters and smaller staging and development clusters that are implemented in hours or minutes. “Organizationally, at the dawn of devops five or six years ago, everyone was talking about the ‘you design, you own’ approach. As a shared platform provider, this means that each application team within Mercedes-Benz has its own Kubernetes cluster,” said Jörg Schüler, team leader at Mercedes-Benz Tech Innovation. “Our goal is to provide an ecosystem and make the application team self-sufficient,” he added. “This ecosystem is supported by self-service principles and API orientation.” This legacy is managed by not one but five separate platform teams. Two of them form a team of 12 engineers focused on the core Kubernetes-as-a-service platform. Here is the team responsible for container security, including database-as-a-service, logging and monitoring as a service, runtime, registry, and image analysis. However, completing such a team is a challenge for the company. “It’s hard to find great Kubernetes experts,” Schüler said. “It’s really helpful to provide education, training, and other services around this platform. Teams of developers help each other through boot camps, training portals, and sandbox environments. It requires a community approach,” added the team leader at Mercedes-Benz Tech Innovation. Golden Path to the Cloud After all these efforts to manage large-scale Kubernetes, Mercedes-Benz Tech Innovation is bringing more and more workloads to the public. We are preparing to move to the cloud. Amazon’s Elastic Kubernetes Service (EKS) eases the cognitive load on our platform and devops team, Peter Müller said, “We don’t know if we’ll choose EKS yet, but we have the same architecture on-site and off-site. That’s why, for the time being, I prefer to opt for EKS myself.” These managed versions of Kubernetes can help ease the burden on the Mercedes-Benz Tech Innovation Platform team, but the application team is still moving to containers and Kubernetes. In this case, a Helm diagram, the “Golden Path”, can provide a solution that accelerates progress in this area.Use it as a template for specific functions, such as identity and access management, so that other teams can do the same. “We need to reduce the cognitive load and provide a gold path and other things as a service to help the team do what they do best: create business value,” Muller said. Each application team is different, and Müller sees his role as providing a secure environment to learn from: “Once these teams have reached sufficient maturity, they can move to the cloud,” he adds. The next Mercedes-Benz Tech Innovation will use in-house sourcing technology to manage some of these golden paths, while others sit in what Müller calls “community stages” awaiting validation. Ideally, this golden path could be codified on a platform similar to the Backstage software catalog platform created by Spotify. “We’re working on a proof-of-concept for a central developer portal that unifies all our services, but we’re not there yet,” says Müller. “Kubernetes management is not difficult for us,” Sabine Wolz, Product Manager at Mercedes-Benz Tech Innovation, said in a speech at KubeCon. Europe. But Perter Müller firmly believes that the learning curve will be mostly application teams, not platform teams. “If you don’t fully manage Kubernetes, it’s hard to manage,” he said. So managing Kubernetes isn’t difficult in the end, because if you choose to manage everything, you have to master it in depth,” said senior expert at Mercedes-Benz Tech Innovation. “Using Kubernetes for application projects is still a challenge. “Using Kubernetes as a devops team is sometimes difficult,” he added. Müller wants his platform team to excel at helping application teams understand the underlying infrastructure without requiring deep expertise. “Some teams are still working on virtual machines and are starting to move to Kubernetes clusters. They have to partition the monolith, understand how transactions are handled, think about asynchronous communication, and understand how Kubernetes works,” he said. “This step is difficult. That’s why we have to help them and not let them sit.”
#MercedesBenz #Kubernetes #clusters #PCWorld

Leave a Comment

Your email address will not be published. Required fields are marked *