Canary deployment em kubernetes com istio: um estudo de caso sobre implantações graduais e gerenciamento de tráfego

Abstract

This paper presents the implementation and analysis of a Canary Deployment strategy using Istio in a Kubernetes cluster. Canary Deployment is a gradual release technique that allows the controlled rollout of new software versions by initially directing a small portion of the traffic to the new version while the majority remains on the stable version. This approach reduces the risks associated with failures and enables the validation of the new version in production before full adoption (NEWMAN, 2015; RICHARDSON, 2018). The study was conducted based on Istio resource configuration, such as Destination Rules, Virtual Services, and Gateways, to manage traffic routing between two versions of an application (v1 and v2). Different traffic distribution scenarios were explored, starting with 90. To monitor and assess the performance of the versions, Prometheus and Kiali tools were used. Prometheus provided detailed metrics on error rates, latency, and throughput, while Kiali offered a graphical view of the application’s topology and traffic flow between services. These tools enabled early problem detection and data-driven decision-making, ensuring system stability.The results demonstrated that the Canary Deployment strategy, combined with Istio, is effective for the safe and gradual deployment of new versions in microservices envi- ronments. The ability to dynamically adjust traffic and perform rapid rollbacks proved crucial in minimizing negative impacts and ensuring system reliability. Additionally, the integration with monitoring tools provided visibility and control over the application’s behavior in real-time.Finally, the paper suggests future directions for improving the implementation, such as automating the Canary Deployment process, integrating with log analysis tools, con- ducting more extensive load testing, and expanding to multi-cluster environments. These improvements have the potential to further increase the efficiency, resilience, and scalabi- lity of the system, solidifying Canary Deployment as an essential practice for managing microservices in production.