Quentin L.

Je contribue à la refonte du système de facturation au sein d’une BU de 90 personnes dédiée au parcours Order-to-Cash. J’assure la conception, le développement et la maintenance de 4 microservices interconnectés, en Go, déployés avec Helm sur Kubernetes, orchestrés via des workflows événementiels Temporal.

Ces services traitent quotidiennement des dizaines de milliers d’événements de facturation, participant à la modernisation d’un module critique historiquement âgé de 10 à 20 ans.

J’ai intégré une équipe passée de 5 à 12 développeurs (+2 lead dev), avec des rituels agiles hebdomadaires (grooming, affinage, revues de code rigoureuses).

Mon rôle implique aussi une forte collaboration inter-équipes pour assurer la cohérence fonctionnelle et technique dans un projet transverse, clef pour la réduction de la dette technique et la fiabilisation de la chaîne de facturation.

» Development of the internal tool "Turbine". Turbine is a containerized application deployment tool integrated into ADEO SERVICES' information service, incorporating various features:

- Scheduled "scale to zero"

- High availability

- Monitoring and alerts based on DataDog

- Secret management based on Vault

- Based on Kubernetes GKE

- Billing

- Automatic deployment via CLI or GitHub Actions

- Fully Golang-developed micro-service management

» Support with on-call duty for the Turbine tool.

» Joint mission with the GKE team within ADEO SERVICES

Within a team of 15 people for the "Managed Kubernetes Service" product, I developed in Go to create and/or improve the Kubernetes operators responsible for the proper delivery of LoadBalancer type services within client clusters, all in a multi-region OpenStack environment.

This involves understanding how Kubernetes works, how to develop within it, understanding reconciliation loops, creating custom objects, and making them "scalable" for a volume of several thousand objects, all with a single operator.

---

1. Study of the existing system and operational maintenance of the multi-region Kubernetes operator enabling the delivery of client load balancers within their managed Kubernetes cluster (approximately 3000 load balancers spread across 5000 Kubernetes clusters). Addressing volume issues that could slow down the delivery process and reduce its reliability. Today, we can deliver a load balancer in each region in under a minute. This is a Kubernetes operator (closely linked to an API also developed in Go) built with Kubebuilder, offering a completely "custom" implementation of the LoadBalancer type service.

2. Study for the redesign of the "Loadbalancer" product to offer the Octavia load balancer provided with Openstack instead of the custom implementation.

To achieve this, we started from the OpenStack cloud-provider-openstack implementation and developed it for integration into the Managed Kubernetes Service product. The 2 load balancer products (the 2 controllers) operate in parallel (2 cloud-controller-managers deployed in client clusters). Later, the Octavia LB will be delivered by default, followed by the decommissioning of the "legacy" infrastructure.