天堂

  :: :: 博问 :: 闪存 :: :: :: 订阅 订阅 :: 管理 ::

故障现象:

中午12点多,一套主从集群的主库因为没有配置大页内存,发布时导致OOM,MYSQL实例重启了,然后MHA发生了切换。切换过程正常。切换后需要把原master配置成新master的slave,在manager.log文件里面找到change master to ....命令,执行后发现复制状态一直停留在connectiong 。名称定:OOM的是M1,挂掉后顶替的是S1.

mysql> show slave status\G
*************************** 1. row ***************************
               Slave_IO_State: Waiting to reconnect after a failed master event read
                  Master_Host: 10.3.171.40
                  Master_User: rep_user
                  Master_Port: 3306
                Connect_Retry: 60
              Master_Log_File: centos-bin.000002
          Read_Master_Log_Pos: 107
               Relay_Log_File: relay-bin.000001
                Relay_Log_Pos: 4
        Relay_Master_Log_File: centos-bin.000002
             Slave_IO_Running: Connecting
            Slave_SQL_Running: Yes
              Replicate_Do_DB: 
          Replicate_Ignore_DB: 
           Replicate_Do_Table: 
       Replicate_Ignore_Table: 
      Replicate_Wild_Do_Table: 
  Replicate_Wild_Ignore_Table: 
                   Last_Errno: 0
                   Last_Error: 
                 Skip_Counter: 0
          Exec_Master_Log_Pos: 107
              Relay_Log_Space: 107
              Until_Condition: None
               Until_Log_File: 
                Until_Log_Pos: 0
           Master_SSL_Allowed: No
           Master_SSL_CA_File: 
           Master_SSL_CA_Path: 
              Master_SSL_Cert: 
            Master_SSL_Cipher: 
               Master_SSL_Key: 
        Seconds_Behind_Master: NULL
Master_SSL_Verify_Server_Cert: No
                Last_IO_Errno: 0
                Last_IO_Error: 
               Last_SQL_Errno: 0
               Last_SQL_Error: 
  Replicate_Ignore_Server_Ids: 
             Master_Server_Id: 2017140

检查错误日志文件,日志如下,提示在S1上找不到master上的binlog文件

160408 12:25:40 [Note] Slave I/O thread: connected to master 'rep_user@10.3.171.40:3306',replication started in log 'centos-bin.000002' at position 107
160408 12:25:40 [ERROR] Error reading packet from server: File '/data2/mysql/centos-bin.000002' not found (Errcode: 2) ( server_errno=29)
160408 12:25:40 [Note] Slave I/O thread: Failed reading log event, reconnecting to retry, log 'centos-bin.000002' at postion 107
160408 12:25:40 [ERROR] Error reading packet from server: File '/data2/mysql/centos-bin.000002' not found (Errcode: 2) ( server_errno=29)
160408 12:26:40 [Note] Slave I/O thread: Failed reading log event, reconnecting to retry, log 'centos-bin.000002' at postion 107
160408 12:26:40 [ERROR] Error reading packet from server: File '/data2/mysql/centos-bin.000002' not found (Errcode: 2) ( server_errno=29)

到S1上去检查,show master status;show master logs可以看到业务数据在写入,POS位置也一直在改变,这里奇怪的是00001文件的大小是0

mysql> show master logs;
+-------------------+-----------+
| Log_name          | File_size |
+-------------------+-----------+
| centos-bin.000001 |         0 |
| centos-bin.000002 | 568661746 |
+-------------------+-----------+
2 rows in set (0.00 sec)

mysql> show master logs;
+-------------------+-----------+
| Log_name          | File_size |
+-------------------+-----------+
| centos-bin.000001 |         0 |
| centos-bin.000002 | 568941034 |
+-------------------+-----------+
2 rows in set (0.00 sec)

mysql> show master logs;
+-------------------+-----------+
| Log_name          | File_size |
+-------------------+-----------+
| centos-bin.000001 |         0 |
| centos-bin.000002 | 569017617 |
+-------------------+-----------+
2 rows in set (0.00 sec)

到data目录查看,却没有找到这2个文件。复制提示也是找不到文件

到这里奇特的现象是:业务正常写数据库,show master status也可以看到有pos位置变化,但是磁盘上没有文件,复制无法建立

[root@GZ_NS_M5_SYNC_mysql_sync1-standby_171.40 ~]# find / -name centos-bin.000002
[root@GZ_NS_M5_SYNC_mysql_sync1-standby_171.40 ~]# 

 

#故障重现

1)正常启动实例,开启binlog,配置复制环境

2)rm 把主库的binlog.index.binlog.0000X删除

3)继续写入数据,pos位置变化

4)从库报错,找不到binlog文件

 

#为什么会出现这样的情况

回想起来这个故障,应该和故障重现的过程是一样的,这套集群3,4个月前搭起来的,在复制正常后,standby的binlog相关文件被删除了,其实删除的整个目录,这个目录专门用来存binlog,relaylog的。删除后搭建复制的时候做change master to,把relay log重建了,但是binlog没有。今天发生了MHA切换,standby变成了master,接受数据写入。MHA里面的filename,pos是连到standby做show master status得到的,但是这些文件已经被删除。所以复制出错。

 

#继续做实验

1)生成binlog.0001后,把binlog.index,binlog.00001都rm后,数据写入,pos逐步变大,当超过1G大小做文件切换,会发生什么?

答:当1写满后做切换,binlog.index没有,拿不到最大的文件ID,那就又从1开始。结论:一直写00001文件

 

2)留下index文件,把00001删除,继续写入,超过1G大小会怎么样?

答:会生成00002文件,这个文件是落地磁盘的正常的binlog文件。

 

#今天出现的故障,如何把events拿出来?

测试下来,如果是statement的,可以通过show master events in xxxx,得到binlog的命令。如果是row格式的,拿不到具体的SQL命令。

 

posted on 2016-04-08 17:42  zuoxingyu  阅读(1161)  评论(0编辑  收藏  举报