Episode #42 07:52 2025-04-30

#042 – Cloud-Native Evolution: Simplifying K8s for Developers with Maxime Veroone (Decathlon)

Maxime Veroone
Staff Engineer, e-commerce division, Decathlon

Listen to the Podcast

Episode Overview

Recorded live at KubeCon London, this short episode of Kubernetes for Humans pairs host Itiel Shwartz with Maxime Veroone, Staff Engineer in Decathlon's e-commerce division. Maxime walks through Decathlon's container journey — from monolithic vendor stacks and Rancher 1 on VMs to a deliberate move onto Kubernetes once shadow K8s started showing up in the org — and what it actually takes to run a platform for 5,000 digital staff. The conversation focuses on developer experience: why hand-written YAML is the wrong interface for most application teams, how Decathlon uses Terraform plus home-grown CI/CD instead of an off-the-shelf portal, and how a strong contribution model with the central platform team beats top-down standardization. Maxime closes on what excites him at the conference, including progressive delivery with Flagger on top of Flux for an e-commerce site where every minute of downtime costs real money.

In this episode we discuss:

  • Decathlon's container path: from monolithic vendor tech to Rancher 1, then to Kubernetes once shadow K8s appeared
  • Running a platform for ~5,000 digital staff and what 'developer experience' actually has to mean at that scale
  • Why Decathlon built its internal platform around Terraform and custom CI/CD instead of Backstage or another off-the-shelf portal
  • The staff-engineer job as a bridge between the central platform team, ops, and front-line software engineers
  • Progressive delivery in e-commerce: using Flagger with Flux so deploys don't take the storefront down

Key Takeaways

1
Shadow Kubernetes inside the org was the real signal to migrate — the team realized their Rancher-plus-automation stack was reinventing K8s primitives.
2
Asking application developers to hand-write ingress and Deployment YAML is the wrong abstraction; the platform's job is to hide that without imposing rigid workflows.
3
Standardizing CI/CD across teams unlocks far more leverage than letting every squad roll its own pipeline.
4
A two-way contribution model with the central platform team — where product teams ship code and feedback into the platform — beats pure top-down standardization.
5
In e-commerce, every minute of downtime is lost revenue, which is why progressive delivery tools like Flagger (paired with Flux) matter more than flashier features.

Itiel Shwartz: Hello everyone and welcome to another episodes of My name is Itiel Shwartz and I’m here at KubeCon London. Today with me we have Maxim, right?

Maxime Veroone: Yeah, that’s me. I’m Maxime Veroone. I work as at Decathlon for as a staff engineer in the e-commerce department.

Itiel Shwartz: That sounds cool. Tell us a bit about yourself. How long are you at Decathlon? Why do you have containers and uh yeah yeah what do you do?

Maxime Veroone: I’ve been working in IT for about 12 years and eight of which being at Decathlon uh I always been at the commerce department and in my past I’ve been working with really old computers back in all the companies 2.2 kernels and RHEL versions older than me etc. the more I went ahead in my career, the more I embraced more modern technologies uh until today where I’m working with Kubernetes, cloud-native technologies, cloud providers, that’s all.

Itiel Shwartz: So maybe share a bit about first of all like why Decathlon needs kubernetes to begin with like what’s the main use case and then maybe talk a bit about the adoption like how did you guys started to adopt?

Maxime Veroone: initially we were working with big monolithic software vendor technologies etc but our developers asked us the platform team to work more with containers when it became a thing and I didn’t want to manage Docker containers on VMs myself no one wants to yeah so I started building something about with Docker Swarm at the time um which enables developers to have resilient workloads scaling workloads not only scaling but already and but it lacked some automation so I built some automation based RunDeck etc around it and we started uh doing some in code so writing some JSON and YAML files to describe what they wanted to deploy on the platform so at some point uh there was starting to be some shadow Kubernetes we dug into that and we just realized that it was the same thing so why not switching like it really enables developer to create value on their software and can standardize and streamline the deployment of their applications and not have to worry too much about how it runs.

Itiel Shwartz: That sounds cool. Like how big is Decathlon engineering? Like is it a big organization?

Maxime Veroone: I believe that today we have about 5,000 people in the uh digital right big company but we spread along a lot of um domains uh which so I’m in e-commerce but there’s a lot more so recently I’ve started working with the internal platform the guys are developing for us and which is really great it enables us to provide some infrastructure

Itiel Shwartz: what are you using like behind the scenes is it Backstage is it something else?

Maxime Veroone: Mainly we build around Terraform with CI/CD everything alone like nothing going off the shelf.

Itiel Shwartz: Yeah. You you describe your YAML file and it becomes infrastructure by the magic of Terraform behind the scene.

Itiel Shwartz: That’s cool. What’s your like biggest like challenge when like in a company like Decathlon with um using of Kubernetes or platform engineering like um what worries you?

Maxime Veroone: What worries me is the complexity we’re operating from developers like in a modern world we’re asking uh software engineers to be QA to be product managers to be and now operators of their own solutions etc. And I feel like as engineers we’re geeks, right? We love the command line, we love files, Git etc. And sometimes uh they struggle to keep up because they have many more things to worry about. So my biggest interest at the moment is to try and create simplicity for developers and also it enables us to have some standardized environments. So for example, CI/CD we used to do everyone does its own CI/CD does whatever they want and that doesn’t help because we deploy a lot of efforts to so I really want to first of all standardize things but most of all I began staff engineer for a few months now and my real goal in that role is to build bridges between the ops guy and the SREs and the software engineers at the front.

Itiel Shwartz: What’s like you I completely agree with you right and we have like the Kubernetes for Humans podcast especially because I do believe that we should lower the barriers of Kubernetes but what do you think is like the the hardest thing for developers in like Decathlon is it the building is it the troubleshooting is it the cost optimization like there are so many different right like things they need to take care of what what do you think is like the most

Maxime Veroone: I mean the for me the difficulty is that Kubernetes and structure is meant to simplify standard etc. deployment of application for developers. But if you ask them to write their own YAML manifest themselves, they’re going to struggle to find the right parameter, have the right parameter, my ingress, my uh Deployment or Service, whatever. It’s so complicated for them. It enables a lot of features. You cannot ask them to be able to do that. So I feel like the main complexity is finding a way to abstract that without imposing too many constraints.

Itiel Shwartz: Yeah. abstract abstraction like leaky abstraction is like one of the hardest thing right like in in in general what what excite you like you know you are here at KubeCon right what cool things have you saw what’s the future like

Maxime Veroone: yeah I love the talks about how companies have been are building platform engineeringing I it’s it’s a big trend right but very big there’s some sense behind that we talk a lot standardization but I mean all of the talks are always about the same thing that listen to your users and that’s something very important to me because at Decathlon I’m a user of platform so it’s all about uh the contribution model and the collaboration with the platform team the central platform so we’ve been doing that great uh in the past years uh we have really great relationship with them and we contribute with our needs sometimes with a few code in the platform we can contribute to you it’s really a win-win situation and because they get the feedback of what we should build how to do in the roadmap etc files

Itiel Shwartz: okay cool anything else you want to say before we wrap up the episode

Maxime Veroone: um yeah if you are building it listen to your users very important and yeah I’m also very excited about Flagger yeah these days because I mean I work in e-commerce like in any minute our website is unavailable it’s a we are not selling stuff and we are not bringing value to our customers so and one of the biggest I always say when you don’t touch anything you don’t break anything.

Itiel Shwartz: Yeah, that’s that’s the reality.

Maxime Veroone: How to ensure that when you touch something you don’t break everything. That’s the whole and Flagger seems to be a real good approach to such especially it engage very well with Flux. It’s made for that. Yeah. And I like Flux too. So good match.

Itiel Shwartz: Okay. Cool. Pleasure having you Max.

Maxime Veroone: Yeah. And enjoy the conference.

Itiel Shwartz: You too. I know.

[Music] Kubernetes for Humans.

This is an AI generated transcript of the conversation

About the Guest

Maxime Veroone
Staff Engineer, e-commerce division, Decathlon
Maxime Veroone is a Staff Engineer in Decathlon's e-commerce division. With about 12 years in IT — eight of them at Decathlon — he started his career working with very old Linux systems and 2.2 kernels before moving steadily toward modern, cloud-native technologies. At Decathlon he helped lead the platform's container journey, starting with Rancher 1 on top of VMs and eventually moving to Kubernetes once the team realized they were rebuilding many of the same primitives. In his current staff role he focuses on building bridges between the central platform team, ops, and software engineers, abstracting Kubernetes complexity without locking developers in.