Worked on enhancing metadata operation reliability and performance for the juicedata/juicefs repository, focusing on high-concurrency scenarios. Addressed transient failures by implementing retry logic for syscall.EBUSY within MySQL and PostgreSQL metadata operations, and refined the parent node update strategy to reduce contention and improve consistency. Leveraged Go for system programming, applying expertise in concurrency, SQL, and performance optimization to enable scalable metadata workloads. These improvements resulted in higher throughput, reduced latency spikes, and greater resilience for large-scale deployments, directly supporting the repository’s performance enhancement initiative and contributing to more robust database-backed metadata management under demanding operational conditions.