基于 Zookeeper 框架实现 MySQL 分布式数据库集群磁

摘要

Data consistency is always based on log replication among MySQL cluster .First ,master database receives io_thread request from slave database ,read log information of specified location and repsond the information to io _thread of slave database .Then ,io_thread of slave databased appends the log information to relay‐log ,sql thread of slave database ana‐lyses relay‐log and executes the sql statements .The MySQL cluster based on log replication lacks unified architecture to real‐ize automated management and needs maintenace for monitoring data consistency ,master‐slave switchover and rejoining clus‐ter after disaster recovery .To realize automated management for MySQL Cluster ,the topic adopts zookeeper to monitor and manage state ,resources ,data consistency and switchover between master and slave databased .%在 MySQL 数据库集群间数据一致性通常是基于日志复制，主数据库接收到来自从库的 io＿thread 请求后，读取日志指定位置之后的日志信息，返回给从库的 io＿thread ，io＿thread 将日志内容添加到 relay‐log ，再由 sql＿thread 解析 re‐lay‐log 并执行。这种基于日志复制数据库集群，缺少统一的框架对 MySQL 集群实现自动化管理，经常需要人工干预集群的维护，如数据一致性监控、主从切换、灾难恢复后加入集群等。为实现 MySQL 集群的自动化管理，论文提出基于 Zoo‐keeper 的管理框架，监控和管理集群的状态、资源、数据一致性和主从切换等。