Connected to a blog, now confirms the 5.6 semi-synchronous replication.
As follows.
Primary library session one query data.
650) this.width=650; "src=" http://s3.51cto.com/wyfs02/M01/6F/8D/wKiom1WfdEqRH0t6AAC3mL88lAY647.jpg "style=" float: none; "title=" screenshot 2015-07-10 pm 3.26.03.png "alt=" Wkiom1wfdeqrh0t6aac3ml88lay647.jpg "/>
Will be stopped from the library IO replication thread, and the main library session continues to insert data, waiting due to synchronous breaking.
650) this.width=650; "src=" http://s3.51cto.com/wyfs02/M02/6F/89/wKioL1Wfdh3i9OenAABfIJX6u2M007.jpg "style=" float: none; "title=" screenshot 2015-07-10 pm 3.26.16.png "alt=" Wkiol1wfdh3i9oenaabfijx6u2m007.jpg "/>
The main library opens session two, queries the data, and discovers that the data for session one has been inserted.
650) this.width=650; "src=" http://s3.51cto.com/wyfs02/M02/6F/8D/wKiom1WfdEugVmy7AACnF2mDSAw246.jpg "style=" float: none; "title=" screenshot 2015-07-10 pm 3.26.24.png "alt=" Wkiom1wfdeugvmy7aacnf2mdsaw246.jpg "/>
Queries the main library Binlog and discovers that the transaction is written to Binlog.
650) this.width=650; "src=" http://s3.51cto.com/wyfs02/M02/6F/8D/wKiom1WfdEzy_vs0AAFuR55ZRIY747.jpg "style=" float: none; "title=" screenshot 2015-07-10 pm 3.27.20.png "alt=" Wkiom1wfdezy_vs0aafur55zriy747.jpg "/>
Summarize:
5.6, after writing Binlog, the transaction is submitted directly to the engine before waiting for the acknowledgement to be returned from the library, where the refresh can see the data. If the main library down at this time, after the master switch, because the thing is not synchronized to from the library, so again refresh the discovery data loss.
This article is from the "Floating Phoenix Year" blog, please be sure to keep this source http://liuzhanbin.blog.51cto.com/10060150/1672920
MySQL5.6 semi-synchronous replication-non-strong consistency