Google Finance Engineering Automates Spanner Dual-Write Migrations

Google Finance Engineering faced the challenge of migrating their legacy data layer to Spanner without taking production services offline, which required rewriting dual-write logic across many Data Access Objects (DAOs). They developed an automated refactoring pipeline using Antigravity CLI in headless mode, significantly accelerating migration and maintaining data parity in staging environments.

Migrating high-throughput production services, especially where financial accuracy is essential, requires a structured approach. The team's strategy involved three distinct phases: historical data backfill, dual-write/dual-read implementation, and automated API verification with byte-for-byte parity checking. Manually implementing these precise changes across numerous DAOs proved slow and error-prone.

To enable reliable code generation, the team standardized their DAO refactoring pattern around a decoupled MutationConverter interface. This approach isolates Spanner schema translation into dedicated converter units, creating a rigid, deterministic contract that an AI coding agent could reliably generate.

For systematic, multi-file code updates across an entire codebase, interactive AI chat interfaces are insufficient. The team built an orchestration script to run Antigravity CLI in headless mode, allowing batch execution, automated verification with self-correction, and overnight processing of code changes to scale their dual-write implementation.

The automated approach yielded clear benefits, including a significant reduction in migration effort, highly reliable data migration with sustained fidelity, and the ability for engineers to focus on higher-value tasks. Key takeaways for other engineers include decoupling schema translation, leveraging headless automation for repetitive refactoring, and integrating AI generation loops directly into build and test systems.