Devsisters Engineering Day: Looking Back at Infra/SRE

At the Devsisters Engineering Day - Infra/SRE, held on February 13, 2025, the team shared the technical problems and solution experiences encountered while stably supporting large-scale game traffic.

The event proceeded with exchanges among participants from various industries, including not only the gaming industry but also hardware manufacturers, combining presentations with networking sessions.

Presentation 1 covered incident response principles and methods. Starting from how to define a service incident, it explained the common response principles applied across all Devsisters development teams, a recommended alarm tiering system, and effective response methods based on real cases.

Alongside the presentation, a related tech blog post was also introduced via QR code.

• Related post: incident-management-principles

Presentation 2 shared the journey of standardizing Kubernetes infrastructure under the name Dalgona. As the number of clusters increased, permission management became more difficult and maintenance costs grew, and there was also the problem that know-how from the central infrastructure organization's migration from Classic Load Balancer (ELB) to Global Accelerator (GA) and Network Load Balancer (NLB) was not well propagated across the organization.

To solve this, they standardized the infrastructure configuration so that company-wide Kubernetes clusters could be operated efficiently, and presented the considerations along the way as well as the future direction for platform engineering.

Presentation 3 introduced the process of building a deployment system based on Argo Workflow that game teams could use without help from the central infrastructure organization. With the existing deployment tool, Docker image tags were fixed and server teams found it difficult to change environment variables directly, making central support essential, and since deployment was possible only through a single chart installation step, only limited separation at the Hook level was possible.

Based on these limitations, Devsisters built its own deployment system, Plan, and established a structure allowing planners and server developers to deploy more independently.

Finally, Devsisters stated that it plans to continue holding engineering days covering various fields beyond Infra/SRE, and also provided recruitment information.