[mercari GEARS 2025] Developer’s Database Operation Issues Revealed through Surveys and Repository Analysis

Transcript

1. Developer’s Database Operation Issues Revealed through Surveys 
 and Repository Analysis
 Tomoyuki Koyama
 Mercari / Database Reliability Engineer

2. Tomoyuki Koyama
 
 
 Started my first career at AWS Japan.
 Joined Mercari in March, 2024.
 Ph.D. student in computer science
 Database Reliability Engineer

3. Background
 • Database Reliability Engineering (DBRE) team ◦ Provide developers with a comfortable DB usage experience ◦ Goal: Developers can quickly and easily solve DB problems • When a developer has a problem regarding DB, the DBRE team supports the developer to solve the problem Developer Microservice Database server App server DBRE Operate Support
4. Investigation
 • DBRE team receives many DB-related support requests including incidents and takes time to handle them • In order to ensure the issue, DBRE team conducted two investigation methods Developers DBRE survey GitHub reopsitories DBRE investigation (1) Developer survey Got 13 responses from different microservice teams (2) Repository analysis Investigated 26 repositories with GitHub MCP server
5. Investigation Results (1) Developer survey (13 responses) →Found performance/capacity problems Votes Options 10 Securely upgrade DBMS versions and scale database servers in/out 9 Backup and restore procedures for database servers 9 Capacity planning for database servers (CPU/RAM/Disk specification determination) and application servers 9 Observability of database servers and database-related application metrics 8 SQL query optimization and index design 8 Troubleshooting guidance and incident analysis support (2) Repository analysis (26 repos) DB scheme management tools are NOT utilized in 32% projects. →Support tool introduction
6. Solution: Database Center
 • A product developed by Google Cloud • Provide a single comprehensive UI for database instances across multiple Google Cloud projects • DBRE is able to find database problems: SQL query optimization / index design Issues Detail
7. Future work
 AI-based problem identification and solving • Database Center supports AI-based problem finding with Gemini • To figure out instances which have resource capacity problems without manually work • AI creates a PR to fix the problem and Developer reviews the PR Developer Can you show resources that have resource capacity problem? AI The following instances have problems regarding resource capacity … I have created a pull request to fix this problem https://github.com/xxx
8. Thank You!