Pika flushdb 改进 #2807
Replies: 2 comments
-
问题简述:因为Slave端消费Binlog线程模型的缘故,原有代码中主从执行flushdb后可能出现主从不一致的现象 背景:Pika Slave在消费一条Binlog时,按事件顺序分为两步,一是Apply Binlog, 也就是将收到的Binlog也追加到自己的BinlogFile中,二是Apply DB, 也即将Binlog中抽取出来的命令写DB。具体地,Slave收到的所有Binlog都会顺序提交到一个固定的Binlog-Worker线程去消费, Case举例:
解决方案:
|
Beta Was this translation helpful? Give feedback.
-
相关 PR #2808 |
Beta Was this translation helpful? Give feedback.
-
记录针对Issue #2708 的修复讨论
Beta Was this translation helpful? Give feedback.
All reactions