Repository navigation
发布说明
- 新功能
Cassandra & ScyllaDB TTL 支持:
新增三种 TTL 迁移模式:source(保留原始 TTL)、target(应用新的固定 TTL)、modify(在原始 TTL 基础上进行时间偏移)。
灵活的数据过滤:
在表映射配置中增加了 where 字段,允许用户自定义过滤条件(SQL/CQL),实现数据的选择性迁移。
全局限速与流量控制:
重构了限速机制,引入全局限速器和突发量(Burst)控制,确保在大规模并发迁移时,整体吞吐量严格符合预期设置。
增强型断点管理:
细化了断点续传配置,支持通过行数阈值(checkpoint_row_threshold)和时间间隔(checkpoint_interval)触发保存,大幅提升了在不稳定网络下的可靠性。 - 性能优化
Cassandra 写入加速:
引入了工作协程池进行并发批处理写入,显著提升了迁移速度。同时针对 ScyllaDB 最佳实践,默认优化了写入批次大小。
指标统计改进:
使用了更加精准的滑动平均算法计算实时迁移速率,并在日志中新增了“剩余行数”和“完成百分比”的动态预估。 - 修复与稳定性
复杂数据类型:
修复了 Cassandra 集合(List/Map/Set)及其嵌套类型的映射问题。
增加了对自定义用户定义类型(UDT)和函数的自动依赖迁移支持。
数据库兼容性:
优化了 MySQL 和 PostgreSQL 的内部处理循环,解决了在高并发下统计信息不准及断点保存可能丢失的问题。
Release Notes
- New Features
TTL Support for Cassandra & ScyllaDB:
Introduced three TTL migration modes: source (replicate as-is), target (apply a new fixed value), and modify (adjust original TTL with a specified offset).
Custom Data Filtering:
Added a new where field in table mappings, allowing users to define custom SQL/CQL filters for selective data transfer.
Global Rate Limiting & Burst Control:
Refactored the throttling mechanism with a global rate limiter and intelligent burst control, ensuring total throughput across all threads stays within limits.
Advanced Checkpointing:
Enhanced reliability with granular checkpoint controls, including row-based thresholds (checkpoint_row_threshold) and time-based intervals (checkpoint_interval). - Performance Enhancements
High-Speed Cassandra Writing:
Implemented a concurrent batch-writing worker pool to significantly increase throughput. Default batch sizes are now optimized per ScyllaDB best practices.
Refined Metrics & Reporting:
Upgraded the rate-calculation algorithm to use a moving average. Logs now include dynamic estimates for "Remaining Rows" and "Completion Percentage." - Bug Fixes & Stability
Complex Type Handling:
Resolved issues with Cassandra collection types (List, Map, Set) and their nested structures.
Added automated migration support for User-Defined Types (UDTs) and Functions.
Database Stability Improvements:
Optimized the internal migration loop for MySQL and PostgreSQL, fixing edge cases in rate-limit integration and checkpoint consistency during high-load operations.