1、切换前背景

社区业务之前部署在腾讯云自建数据库上，因自建数据库不能动态扩容，需要把数据库迁移到商业数据库上

MySQL-A，自从部署之后一直是Master，未进行过主从切换，GTID: “aaaaaaaa-aaaa-aaaa-aaaa-aaaa20190409:1-535395706”
Canal-Server ，从数据库部署开始，就进行数据同步相关工作
MySQL-B, 为进行数据库迁移工作开始进行主库的同步操作

2、主从迁移操作

DBA侧操作
1.我先改域名指向
2.你们重启服务，重启后新库指向只读状态，不可写
3.我断开同步 数据库可写

3、Canal-Server 变化

迁移操作后，不是简单的数据主从复制

• 新集群和老集群 结构是不一样的
• 老集群产生的 binlog，在新集群中是查不到的
• canal-server 重连后，指向了新的master集群，原来存储的gtid对应的binlog已经找不到了

(后来canal-server 重连后，用保存的gtid(gtid=aaaaaaaa-aaaa-aaaa-aaaa-aaaa20190409:1-535395706) 去新master拉取binlog ，显示master has purged了gtid)

(1)、完成主从迁移操作后，主从关系发生变化
MySQL-B，变成新的Master，Gtid 变为 “aaaaaaaa-aaaa-aaaa-aaaa-aaaa20190409:1-535395706,ee60863f-fbe9-11eb-8e87-1c34da5039bc:1-8”
(2)、主从切换时，原Master MySQL-A 不能连接，连接中断,Canal-Server 会有对应日志输出 原日志信息

2021-09-07 00:08:38.011 [destination = yiCheForumCanal , address = mysqlforum.service.yctxy/10.20.26.51:3306 , EventParser] WARN  c.a.o.canal.parse.inbound.mysql.dbsync.DirectLogFetcher - Received EOF packet from server, apparent master disconnected. It's may be duplicate slaveId , check instance config
2021-09-07 00:08:54.204 [destination = yiCheForumCanal , address = mysqlforum.service.yctxy/10.20.26.51:3306 , EventParser] WARN  c.a.o.c.p.inbound.mysql.rds.RdsBinlogEventParserProxy - ---> begin to find start position, it will be long time for reset or first position
2021-09-07 00:08:54.218 [destination = yiCheForumCanal , address = mysqlforum.service.yctxy/10.20.26.51:3306 , EventParser] WARN  c.a.o.c.p.inbound.mysql.rds.RdsBinlogEventParserProxy - ---> find start position successfully, EntryPosition[included=false,journalName=binlog.000528,position=796262297,serverId=2651,gtid=aaaaaaaa-aaaa-aaaa-aaaa-aaaa20190409:1-535395706,timestamp=1630944273000] cost : 14ms , the next step is binlog dump
2021-09-07 00:08:54.223 [destination = yiCheForumCanal , address = mysqlforum.service.yctxy/10.20.26.51:3306 , EventParser] ERROR c.a.o.canal.parse.inbound.mysql.dbsync.DirectLogFetcher - I/O error while reading from client socket
java.io.IOException: Received error packet: errno = 1236, sqlstate = HY000 errmsg = The slave is connecting using CHANGE MASTER TO MASTER_AUTO_POSITION = 1, but the master has purged binary logs containing GTIDs that the slave requires.
        at com.alibaba.otter.canal.parse.inbound.mysql.dbsync.DirectLogFetcher.fetch(DirectLogFetcher.java:102) ~[canal.parse-1.1.5.jar:na]
        at com.alibaba.otter.canal.parse.inbound.mysql.MysqlConnection.dump(MysqlConnection.java:263) [canal.parse-1.1.5.jar:na]
        at com.alibaba.otter.canal.parse.inbound.AbstractEventParser$1.run(AbstractEventParser.java:255) [canal.parse-1.1.5.jar:na]
        at java.lang.Thread.run(Thread.java:745) [na:1.8.0_111]

• Canal-Server 发现原Master 主库连接异常，”Received EOF packet from server, apparent master disconnected”
• 可以正常连接，Canal-Server 重新开始工作流程
• Canal-Serer 是模仿MySQL Slave, 只能想Master发送命令请求，不能获取最新Master MySQL-B 的GTID，就拿着Canal-Server机制保存的GTID，去向新的Master MySQL-B 服务器获取binlog日志，然后错误异常”

…but the master has purged binary logs containing GTIDs that the slave requires.”，Master服务无法获取GTID对应的binlog日志

4、Canal操作

(1)、DBA通过，获取当前主库开始的GTID

(2)、Canal Server 修改instance.xml 里面instance.master.gtid=
(3)、重新启动，发现新错误

(4)、修改tsdb=false，重启，正常消费

5、Canal-Server反思

(1)、Canal-Server 感知到主从切换时，没有向当前Master: MySQL-B 发送获取GTID的命令，没有重新获取最新的GTID，而是以自己保存的GTID 进行binlog拉取，导致异常报错，后续无法进行

6、问题

出现 apparent master disconnected日志后，修改流程

疑问点:
1、canal-server 停止一段后，重新启动，如何保证GTID还能正常进行获取binlog，purged 怎么办
2、主从复制，这么获取gtid，是否正确，可保证数据不丢失？
3、目前canal 出现连接中断，再次重连后，用canal-server中保存gtid 进行binlog拉取，gtid对应位点不存在，则报错误(errmsg = The slave is connecting using CHANGE MASTER TO MASTER_AUTO_POSITION = 1, but the master has purged binary logs containing GTIDs that the slave requires.)
4、