<span style="font-size:16px;"><strong>一 前言</strong></span><br /> <span style="font-size:12px;font-family:NSimSun;">&nbsp; MySQL 的主从复制作为一项高可用特性，用于将主库的数据同步到从库，在维护主从复制数据库集群的时候，作为专职的MySQL DBA，笔者相信大多数人都会遇到</span><span style="font-size:12px;font-family:NSimSun;">“</span><span style="white-space:normal;font-size:12px;font-family:NSimSun;">Got fatal error 1236 from master when reading data from binary log</span><span style="font-size:12px;font-family:NSimSun;">” 这类的报错/报警。本文整理了常见的几种 error 1236 报错，并给出相应的解决方法，有所不足之处，</span><span style="font-size:12px;font-family:NSimSun;">当然也希望各位读者朋友指正。</span><br /> <br /> <span style="font-size:16px;"><strong>二 常见的error 1236 报错</strong></span><br /> <strong><span style="font-size:14px;font-family:SimSun;">2.1 logevent超过</span></strong><span style="white-space:normal;"><strong><span style="font-size:14px;font-family:SimSun;">max_allowed_packet 大小<br /> <div id="codeText" class="codeText"> <ol style="margin:0 1px 0 0;padding:5px 0pt;" start="1" class="dp-css none_number"> <li> <span style="color:#000000;">Got fatal error 1236 from master when reading data from binary log<span style="color:#0000CC;">:</span> <span style="color:#FF00FF;">'log event entry exceeded max_allowed_packet; Increase max_allowed_packet on master; the start event position from '</span>mysql<span style="color:#0000CC;">-</span>bin<span style="color:#0000CC;">.</span>006730<span style="color:#FF00FF;">' at 290066246, the last event was read from '/</span>u01/my3309/log/mysql<span style="color:#0000CC;">-</span>bin<span style="color:#0000CC;">.</span>006730</span> </li> </ol> </div> </span></strong></span><span style="font-size:12px;"> <span style="font-size:14px;font-family:SimSun;">【<strong>原因</strong>】</span></span><br /> <span style="font-size:12px;">&nbsp; &nbsp;此类报错和max_allowed_packet相关。首先max_allowed_packet控制着主从复制过程中，一个语句产生的二进制binlog event大小，它的值必须是1024的倍数 。出现此类错误的常见原因是</span><br /> <span style="font-size:12px;font-family:SimSun;">&nbsp;1 该参数在主备库的配置大小不一样，主库的配置值大于从库的配置值。 从主库传递到备库的binlog event大小超过了主库或者备库的max_allowed_packet大小。</span><br /> <span style="font-size:12px;font-family:SimSun;">&nbsp;2 主库有大量数据写入时，比如在主库上执行&nbsp;laod data，insert into .... select 语句，产生大事务。</span><br /> <span style="font-size:12px;font-family:SimSun;"> 当主库向从库传递一个比从库的max_allowed_packet 大的packet ，从库接收该packet失败，并报 “log event entry exceeded max_allowed_packet“。</span><br /> <span style="font-size:12px;"> <span style="font-size:14px;">【<strong>如何解决</strong>】</span></span><br /> <span style="font-size:12px;"> &nbsp;需要确保主备配置一样，然后尝试调大该参数的值。</span><span style="font-size:12px;"><br /> <div id="codeText" class="codeText"> <ol style="margin:0 1px 0 0;padding:5px 0pt;" start="1" class="dp-css none_number"> <li> <span style="color:#000000;">set global max_allowed_packet <span style="color:#0000CC;">=</span>1<span style="color:#0000CC;">*</span>1024<span style="color:#0000CC;">*</span>1024<span style="color:#0000CC;">*</span>1024<span style="color:#0000CC;">;</span><br /> </span> </li> <li> stop slave<span style="color:#0000CC;">;</span><br /> </li> <li> start slave </li> </ol> </div> </span><span style="font-size:12px;">另外，5.6 版本中的 <a href="http://dev.mysql.com/doc/refman/5.6/en/replication-options-slave.html#sysvar_slave_max_allowed_packet" target="_blank">slave_max_allowed_packet_size</a> 参数控制slave 可以接收的最大的packet 大小，该值通常大于而且可以覆盖 max_allowed_packet 的配置, 进而减少由于上面的问题导致主从复制中断。</span><br /> <span style="font-size:12px;font-family:SimSun;"><strong><span style="font-size:14px;"><br /> 2.2 slave 在主库找不到binlog文件</span></strong>&nbsp;</span><br /> <div id="codeText" class="codeText"> <ol style="margin:0 1px 0 0;padding:5px 0pt;" start="1" class="dp-css none_number"> <li> <span style="color:#000000;">Got fatal error 1236 from master when reading data from binary log<span style="color:#0000CC;">:</span></span> </li> </ol> </div> <span style="font-size:12px;font-family:SimSun;"> <span style="font-size:14px;">【<strong>原因</strong>】</span></span><br /> <span style="font-size:12px;font-family:SimSun;">&nbsp;该错误发生在从库的io进程从主库拉取日志时，发现主库的mysql_bin.index文件中第一个文件不存在。出现此类报错可能是由于你的slave 由于某种原因停止了好长一段是时间，当你重启slave 复制的时候，在主库上找不到相应的binlog ,会报此类错误。或者是由于某些设置主库上的binlog被删除了，导致从库获取不到对应的binglog file。</span><br /> <span style="font-size:12px;font-family:SimSun;"> <span style="font-size:14px;">【<strong>如何解决</strong>】</span></span><br /> <span style="font-size:12px;font-family:SimSun;"> &nbsp;1 为了避免数据丢失，需要重新搭建slave 。</span><br /> <span style="font-size:12px;font-family:SimSun;"> &nbsp;2 注意主库binlog的清理策略，选择基于时间过期的删除方式还是基于空间利用率的删除方式。</span><br /> <span style="font-size:12px;font-family:SimSun;">&nbsp; 不要使用rm -fr 命令删除binlog file，这样不会同步修改mysql_bin.index 记录的binlog 条目。在删除binlog的时候确保主库保留了从库 show slave status 的Relay_Master_Log_File对应的binlog file。</span><br /> <span style="font-family:SimSun;"><strong><span style="font-size:14px;"><br /> 2.3 主库空间问题，</span></strong><strong><span style="font-size:14px;">日志被截断</span></strong></span><span style="font-family:SimSun;"><br /> <div id="codeText" class="codeText"> <ol style="margin:0 1px 0 0;padding:5px 0pt;" start="1" class="dp-css none_number"> <li> <span style="color:#000000;">Got fatal error 1236 from master when reading data from binary log<span style="color:#0000CC;">:</span> <span style="color:#FF00FF;">'binlog truncated in the middle of event; consider out of disk space on master; the start event position from '</span>mysql<span style="color:#0000CC;">-</span>bin<span style="color:#0000CC;">.</span>006730<span style="color:#FF00FF;">' at 290066434, the last event was read from '</span><span style="color:#FF00FF;">/</span>u01/my3309/log/mysql<span style="color:#0000CC;">-</span>bin<span style="color:#0000CC;">.</span>006730</span> </li> </ol> </div> </span> <span style="font-size:12px;font-family:SimSun;"> <span style="font-size:14px;">【<strong>原因</strong>】</span></span><br /> <span style="font-size:12px;font-family:SimSun;">&nbsp;该错误和主库的空间问题和sync_binlog配置有关，当主库 sync_binlog=N不等于1且磁盘空间满时,MySQL每写N次binary log，系统才会同步到磁盘，但是由于存储日志的磁盘空间满而导致MySQL 没有将日志完全写入磁盘，binlog event被截断。slave 读取该binlog file时就会报错"binlog truncated in the middle of event;"<br /> &nbsp;<span style="white-space:normal;font-family:SimSun;">当sync_binlog 的默认值是0，像操作系统刷其他文件的机制一样，MySQL不会同步到磁盘中去而是依赖操作系统来刷新binary log。</span><br style="white-space:normal;" /> <span style="white-space:normal;font-family:SimSun;">&nbsp;当sync_binlog =N (N&gt;0) ，MySQL 在每写 N次 二进制日志binary log时，会使用fdatasync()函数将它的写二进制日志binary log同步到磁盘中去。</span></span><br /> <span style="font-size:12px;font-family:SimSun;"> <span style="font-size:14px;">【<strong>如何解决</strong>】</span></span><br /> <span style="font-size:12px;font-family:SimSun;">&nbsp;在从库重新指向到主库下一个可用的binlog file 并且从binlog file初始化的位置开始</span><br /> <div id="codeText" class="codeText"> <ol style="margin:0 1px 0 0;padding:5px 0pt;" start="1" class="dp-css none_number"> <li> <span style="color:#000000;">stop slave<span style="color:#0000CC;">;</span><br /> </span> </li> <li> change master to master_log_file<span style="color:#0000CC;">=</span><span style="color:#FF00FF;">'mysql-bin.006731'</span><span style="color:#0000CC;">,</span> master_log_pos<span style="color:#0000CC;">=</span>4<span style="color:#0000CC;">;</span><br /> </li> <li> start slave; </li> </ol> </div> <strong><span style="font-size:14px;">2.4 主库异常断电,从库读取错误的position</span></strong><br /> <div id="codeText" class="codeText"> <ol style="margin:0 1px 0 0;padding:5px 0pt;" start="1" class="dp-css none_number"> <li> <span style="color:#000000;">120611 20<span style="color:#0000CC;">:</span>39<span style="color:#0000CC;">:</span>38 <span style="color:#0000CC;">[</span>ERROR<span style="color:#0000CC;">]</span> Error reading packet from server<span style="color:#0000CC;">:</span> Client requested master to start replication from impossible position <span style="color:#0000CC;">(</span> server_errno<span style="color:#0000CC;">=</span>1236<span style="color:#0000CC;">)</span> <br /> </span> </li> <li> 120611 20<span style="color:#0000CC;">:</span>39<span style="color:#0000CC;">:</span>38 <span style="color:#0000CC;">[</span>ERROR<span style="color:#0000CC;">]</span> Slave I/O<span style="color:#0000CC;">:</span> Got fatal error 1236 from master when reading data from binary log<span style="color:#0000CC;">:</span> <span style="color:#FF00FF;">'Client requested master to start replication from impossible position'</span><span style="color:#0000CC;">,</span> Error_code<span style="color:#0000CC;">:</span> 1236<br /> </li> <li> 120611 20<span style="color:#0000CC;">:</span>39<span style="color:#0000CC;">:</span>38 <span style="color:#0000CC;">[</span>Note<span style="color:#0000CC;">]</span> Slave I/O thread exiting<span style="color:#0000CC;">,</span> read up to log <span style="color:#FF00FF;">'mysql-bin.000143'</span><span style="color:#0000CC;">,</span> position 664526789 </li> </ol> </div> <span style="font-size:12px;font-family:SimSun;"> <strong><span style="font-size:14px;">【原因】</span></strong></span><br /> <span style="font-size:12px;font-family:SimSun;"> &nbsp;该问题也是和sync_binlog=N不等于1有关，多出现在主机异常crash ，比如磁盘损坏，raid 卡损坏，或者主机异常掉电导致binlog 未及时同步到磁盘。从库读取了主库binlog file中的不存在的binlog position ，一般比binlogfile 的end position 的值还要大。</span><br /> <span style="font-size:12px;font-family:SimSun;"> <span style="font-size:14px;">【<strong>如何解决</strong>】</span></span><br /> <span style="font-size:12px;font-family:SimSun;"> 1 在从库重新指向到主库下一个可用的binlog file 并且从binlog file初始化的位置开始</span><span style="font-family:SimSun;"><br /> <div id="codeText" class="codeText"> <ol style="margin:0 1px 0 0;padding:5px 0pt;" start="1" class="dp-css none_number"> <li> <span style="color:#000000;">stop slave<span style="color:#0000CC;">;</span><br /> </span> </li> <li> change master to master_log_file<span style="color:#0000CC;">=</span><span style="color:#FF00FF;">'mysql-bin.000144'</span><span style="color:#0000CC;">,</span> master_log_pos<span style="color:#0000CC;">=</span>4<span style="color:#0000CC;">;</span><br /> </li> <li> start slave; </li> </ol> </div> </span><span style="font-size:12px;font-family:SimSun;">2 主备库设置 sync_binlog=1，但是设置为1的时候，会带来性能下降。&nbsp;</span><br /> <br /> <span style="font-size:16px;font-family:SimSun;"><strong>三 相关阅读</strong></span><br /> <span style="font-size:12px;font-family:SimSun;"> 1 <a href="http://dev.mysql.com/doc/refman/5.5/en/server-system-variables.html#sysvar_max_allowed_packet" target="_blank">max_allowed_packet 官方介绍</a></span><br /> <span style="font-size:12px;font-family:SimSun;"> 2 <a href="https://www.percona.com/blog/2014/10/08/mysql-replication-got-fatal-error-1236-causes-and-cures/" target="_blank">Percona MySQL的特性</a>&nbsp;max_binlog_files &nbsp;&nbsp;</span><br /> <span style="font-size:12px;font-family:SimSun;"> 3 <a href="http://blog.itpub.net/22664653/viewspace-1063134/" target="_blank">sync_binlog innodb_flush_log_at_trx_commit 浅析</a> &nbsp;</span><br /> <div style="white-space:nowrap;"> <span style="font-size:12px;font-family:NSimSun;">4</span><a href="http://blog.itpub.net/22664653/viewspace-732503/" target="_blank"> 《<span style="font-size:12px;font-family:SimSun;">主库异常断电,从库读取错误的position 案例》</span></a><strong style="white-space:normal;"><span style="font-size:14px;">&nbsp;</span></strong><span style="font-size:14px;"><br /> <a href="https://www.percona.com/blog/2014/05/14/max_allowed_packet-and-binary-log-corruption-in-mysql/" target="_blank"><span style="font-family:SimSun;font-size:12px;">5 《max_allowed_packet and binary log corruption in MySQL》</span><span style="font-family:SimSun;font-size:12px;"></span></a></span> </div>