Context
Goal / Problem
Due to the company's age and history of pivots, the backend services contained a significant amount of technical debt in the form of obsolete, unused API endpoints. The manual process to verify usage involved multiple engineers running individual queries in AWS CloudWatch against AWS ELB access logs for every potential endpoint. This manual verification process was slow and cumbersome, taking a multi-engineer team approximately two weeks per verification cycle to confirm which endpoints could be safely retired.
Role
Primary developer and architect for the custom automation tool. My contributions included:
- Conceiving the solution to automate the usage verification process using a dedicated CLI.
- Designing the logic to ingest the router table from our Go services.
- Developing the tool in Go using the AWS SDK for Go to interface with S3.
- Implementing the parsing and querying logic to process AWS ELB access logs retrieved from S3 and count endpoint usage.
Tech Stack
- Frontend: N/A (Command Line Interface - CLI)
- Backend: Go (for CLI tool)
- Database: N/A (data source is AWS ELB access logs stored in S3)
- Infra / DevOps: AWS S3, AWS ELB, AWS SDK for Go
Architecture Overview
The tool is a Go-based Command Line Interface (CLI) application.
- Input: The CLI accepts parameters defining the target Go service's router table and the time period for analysis.
- Extraction: It automatically reads and parses the defined router table to get a list of all known API endpoints.
- Data Retrieval: Using the AWS SDK for Go, the tool programmatically accesses and retrieves the raw AWS ELB access logs stored in AWS S3 for the specified time window.
- Processing: The Go tool processes the logs, tallying the usage count for each known endpoint.
- Output: It generates a report displaying a usage count for all endpoints or simply listing those endpoints confirmed to have zero usage (obsolete).
Implementation Highlights
- Developed a performant, custom CLI tool in Go to replace a complex manual workflow.
- Utilized the AWS SDK for Go to programmatically access and efficiently parse raw, potentially large, ELB log files stored in S3.
- Created a core routine that efficiently maps ELB log request paths to the actual Go service router table definitions to ensure accurate usage attribution.
- Reduced the runtime of the usage verification process from a two-week multi-engineer effort to a 30-minute automated job.
Challenges & Decisions
- Decision: Built a custom Go CLI to parse raw S3 logs instead of relying on specialized CloudWatch querying.
- Rationale: While CloudWatch was the initial manual tool, its query language required per-endpoint effort. Building a custom tool in Go allowed for programmatic ingestion of the entire router map and processing a batch of logs simultaneously, providing massive efficiency gains that no manual query system could match.
Outcome / Impact
- Eliminated weeks of manual engineering effort per verification cycle, redirecting valuable engineer time to new feature development.
- Reduced the total time-to-verification for a six-month usage period from a two-week sprint to approximately 30 minutes.
- Enabled the safe removal of 20+ obsolete API endpoints, significantly reducing technical debt and simplifying future maintenance and security reviews.
Lessons / Next Steps
- Gained deep practical experience with the AWS SDK for Go and large-scale log processing/parsing.
- Learned that investment in simple, targeted internal tooling can yield exponential returns in efficiency and technical debt reduction.
- Next step would be to integrate this tool into the CI/CD pipeline to automatically flag zero-usage routes prior to every major release.