mariadb 主从复制

一、安装依赖包

yum install -y net-tools
yum install -y wget

# 切换aliyun的yum源
cd /etc/yum.repos.d/
mv CentOS-Base.repo CentOS-Base.repo.bak 
wget -O /etc/yum.repos.d/CentOS-Base.repo http://mirrors.aliyun.com/repo/Centos-7.repo 
# 重建源数据缓存 
yum clean all
yum makecache 

yum -y install vim-enhanced wget net-tools telnet vim lrzsz ntp
yum -y install lshw pciutils gdisk system-storage-manager 
yum -y install bash-completion zip unzip bzip2 tree tmpwatch pinfo man-pages
yum -y install nano vim-enhanced tmux screen
yum -y install net-tools psmisclsof sysstat
yum -y install yum-plugin-security yum-utils createrepo
yum -y install get wget curl eliks lynx lftp mailx mutt reync 
yum -y install libaio make cmake gcc-c++ gcc zib zlib-devel open openssl-devel pcre pcre-devel

二、创建用户

useradd -M -r -s  /bin/nologin mysql
echo "mysql"|passwd --stdin mysql

三、下载安装

cd /opt/
wget http://mariadb.mirror.nucleus.be//mariadb-10.5.0/bintar-linux-systemd-x86_64/mariadb-10.5.0-linux-systemd-x86_64.tar.gz
tar -zxvf mariadb-10.5.0-linux-systemd-x86_64.tar.gz -C /usr/local/
cd /usr/local/
ln -s mariadb-10.5.0-linux-systemd-x86_64/ mysql
echo "PATH=/usr/local/mysql/bin:$PATH" > /etc/profile.d/mysql.sh

四、配置文件

cat >/etc/my.cnf<<"EOF"
[client]
port    = 3306
socket    = /data/mysql/data/mysql.sock
default-character-set=utf8mb4

[mysql]
prompt="\ \R:\m:\s [\d]> "
no-auto-rehash
default-character-set=utf8mb4

[mysqld]
user    = mysql
port    = 3306
extra_port=13306
socket  = /data/mysql/data/mysql.sock
basedir= /usr/local/mysql
datadir= /data/mysql/data
log-error=/data/mysql/logs/mysql-error.log
pid-file=/data/mysql/data/mariadb.pid
character-set-server = utf8mb4
skip_name_resolve = 1
open_files_limit = 65535
back_log = 1024
max_connections = 512
max_connect_errors = 1000000
table_open_cache = 1024
table_definition_cache = 1024
thread_stack = 512K
external-locking = FALSE
max_allowed_packet = 32M
sort_buffer_size = 4M
join_buffer_size = 4M
thread_cache_size = 768
query_cache_size = 0
query_cache_type = 0
interactive_timeout = 600
wait_timeout = 600
tmp_table_size = 32M
max_heap_table_size = 32M
slow_query_log = 1
slow_query_log_file = /data/mysql/logs/slow.log
long_query_time = 0.1
log_queries_not_using_indexes =1
min_examined_row_limit = 100
log_slow_admin_statements = 1
log_slow_slave_statements = 1
server-id = 3306
log-bin = /data/mysql/logs/mybinlog
sync_binlog = 1
binlog_cache_size = 4M
max_binlog_cache_size = 2G
max_binlog_size = 1G
expire_logs_days = 30
log_slave_updates
binlog_format = row
binlog_checksum = 1
relay_log_recovery = 1
relay-log-purge = 1
key_buffer_size = 32M
read_buffer_size = 8M
read_rnd_buffer_size = 4M
bulk_insert_buffer_size = 64M
myisam_sort_buffer_size = 128M
myisam_max_sort_file_size = 10G
myisam_repair_threads = 1
lock_wait_timeout = 3600
innodb_thread_concurrency = 0
innodb_sync_spin_loops = 100
innodb_spin_wait_delay = 30

transaction_isolation = REPEATABLE-READ
#innodb_additional_mem_pool_size = 16M
innodb_buffer_pool_size = 5734M
innodb_buffer_pool_instances = 8
innodb_data_file_path = ibdata1:1G:autoextend
innodb_flush_log_at_trx_commit = 1
innodb_log_buffer_size = 32M
innodb_log_file_size = 2G
innodb_log_files_in_group = 2

# 根据您的服务器IOPS能力适当调整
# 一般配普通SSD盘的话，可以调整到 10000 - 20000
# 配置高端PCIe SSD卡的话，则可以调整的更高，比如 50000 - 80000
innodb_io_capacity = 8000
innodb_write_io_threads = 8
innodb_read_io_threads = 8
innodb_purge_threads = 1
innodb_open_files = 65535
innodb_max_dirty_pages_pct = 50
innodb_flush_method = O_DIRECT
#innodb_checksums = 1
#innodb_file_format = Barracuda
#innodb_file_format_max = Barracuda
innodb_lock_wait_timeout = 10
innodb_rollback_on_timeout = 1
innodb_print_all_deadlocks = 1
innodb_file_per_table = 1
innodb_stats_on_metadata = 0
symbolic-links=0
innodb_status_file = 1

#performance_schema
performance_schema = 1

# 字符集设定utf8mb4
character-set-server=utf8mb4
collation-server=utf8mb4_unicode_ci
init-connect=‘SET NAMES utf8mb4‘

# 优化
optimizer_switch = "mrr=on,mrr_cost_based=on,mrr_sort_keys=on"
deadlock_search_depth_short = 3 
deadlock_search_depth_long = 10 
deadlock_timeout_long = 10000000 
deadlock_timeout_short = 5000
slave-parallel-threads=8

# gtid
gtid_strict_mode=1
wsrep_gtid_mode=1

[mysqld_safe]
nice=-19
open-files-limit=65535

[mysqldump]
quick
max_allowed_packet = 64M

EOF

SERVIER_ID=`date +%S`
sed -i "s/server-id = 3306/server-id = 3306"${SERVIER_ID}"/" /etc/my.cnf

my.cnf

五、修改权限

mkdir -p /data/mysql/{data,logs}
chown mysql.mysql -R /data/mysql
chown mysql.mysql -R /usr/local/mariadb-10.5.0-linux-systemd-x86_64
chown mysql.mysql -R /usr/local/mysql

六、初始化

# 初始化数据
cd /usr/local/mysql
./scripts/mysql_install_db  --defaults-file=/etc/my.cnf --user=mysql --basedir=/usr/local/mysql --datadir=/data/mysql/data

七、启用关闭

# 启动mariadb
ls /usr/local/mysql/support-files/systemd/
cp support-files/systemd/mariadb.service /lib/systemd/system/
systemctl daemon-reload
systemctl start mariadb

# 关闭maridb
systemctl stop mariadb

八、登录

# 登录
[ mysql]# mysql
Welcome to the MariaDB monitor.  Commands end with ; or \g.
Your MariaDB connection id is 5
Server version: 10.5.0-MariaDB-log MariaDB Server

Copyright (c) 2000, 2018, Oracle, MariaDB Corporation Ab and others.

Type ‘help;‘ or ‘\h‘ for help. Type ‘\c‘ to clear the current input statement.

 10:36:  [(none)]> show databases;
+--------------------+
| Database           |
+--------------------+
| information_schema |
| mysql              |
| performance_schema |
| test               |
+--------------------+
4 rows in set (0.000 sec)

 10:36:  [(none)]> delete from mysql.user where user=‘‘;
 10:36:  [(none)]> drop database test;
# 创建管理员
grant all privileges on *.* to ‘root‘@‘127.0.0.1‘ identified by ‘root‘ with grant option;
# 登录超级管理端口
mysql -h 127.0.0.1 -uroot -p -P13306

九、主从搭建

# 创建同步用户

# mariadb搭建主从
CREATE USER ‘repl‘@‘%‘ identified by ‘replpwd‘; 
GRANT REPLICATION SLAVE ON *.* TO ‘repl‘@‘%‘

# 主库操作

# 主库上导出全备数据
[ opt]# mysqldump --all-databases --gtid=on --master-data=2 --single-transaction >full_backup.sql
[ ~]# head -35 /opt/full_backup.sql |grep -Ei ‘change|gtid_slave_pos‘
-- CHANGE MASTER TO MASTER_LOG_FILE=‘mybinlog.000007‘, MASTER_LOG_POS=11792;
-- CHANGE MASTER TO MASTER_USE_GTID=slave_pos;
-- SET GLOBAL gtid_slave_pos=‘0-330625-63‘;
[ ~]#

# 从库操作

# 在从库上操作

reset master ;
SET GLOBAL gtid_slave_pos=‘0-330625-63‘;
CHANGE MASTER TO
  MASTER_HOST=‘192.168.142.136‘,
  MASTER_USER=‘repl‘,
  MASTER_PASSWORD=‘replpwd‘,
  MASTER_PORT=3306,
  MASTER_USE_GTID = slave_pos;
 

start slave;
show slave status\G;

十、模拟故障操作

#1、在从库上操作，然后在主库上操作，模拟同步故障
 15:38:  [test]> set sql_log_bin=0;          
Query OK, 0 rows affected (0.000 sec)

 15:39:  [test]> delete from t1 where id = 1;
Query OK, 1 row affected (0.002 sec)

 15:40:  [test]> select * from t1 limit 1;   
+----+---------+---------+------------+---------+--------+----------+--------+
| id | pay_min | pay_max | grade_type | subject | period | discount | price  |
+----+---------+---------+------------+---------+--------+----------+--------+
|  2 |      21 |      59 |          1 |       0 |      0 |       90 | 195.00 |
+----+---------+---------+------------+---------+--------+----------+--------+
1 row in set (0.000 sec)
 15:43:  [test]> set sql_log_bin=1;
Query OK, 0 rows affected (0.000 sec)

# 主库
 15:35:  [test]> delete from t1 where id=1;
Query OK, 1 row affected (0.027 sec)

# 从库，查看同步状态
 15:45:  [test]> show slave status\G;
*************************** 1. row ***************************
                Slave_IO_State: Waiting for master to send event
                   Master_Host: 192.168.142.136
                   Master_User: repl
                   Master_Port: 3306
                 Connect_Retry: 60
               Master_Log_File: mybinlog.000007
           Read_Master_Log_Pos: 23722
                Relay_Log_File: mariadb-relay-bin.000002
                 Relay_Log_Pos: 12386
         Relay_Master_Log_File: mybinlog.000007
              Slave_IO_Running: Yes
             Slave_SQL_Running: No
               Replicate_Do_DB: 
           Replicate_Ignore_DB: 
            Replicate_Do_Table: 
        Replicate_Ignore_Table: 
       Replicate_Wild_Do_Table: 
   Replicate_Wild_Ignore_Table: 
                    Last_Errno: 1032
                    Last_Error: Could not execute Delete_rows_v1 event on table test.t1; Can‘t find record in ‘t1‘, Error_code: 1032; handler error HA_ERR_END_OF_FILE; the event‘s master log mybinlog.000007, end_log_pos 23691
                  Skip_Counter: 0
           Exec_Master_Log_Pos: 23480
               Relay_Log_Space: 12939
               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: 1032
                Last_SQL_Error: Could not execute Delete_rows_v1 event on table test.t1; Can‘t find record in ‘t1‘, Error_code: 1032; handler error HA_ERR_END_OF_FILE; the event‘s master log mybinlog.000007, end_log_pos 23691
   Replicate_Ignore_Server_Ids: 
              Master_Server_Id: 330625
                Master_SSL_Crl: 
            Master_SSL_Crlpath: 
                    Using_Gtid: Slave_Pos
                   Gtid_IO_Pos: 0-330625-68
       Replicate_Do_Domain_Ids: 
   Replicate_Ignore_Domain_Ids: 
                 Parallel_Mode: conservative
                     SQL_Delay: 0
           SQL_Remaining_Delay: NULL
       Slave_SQL_Running_State: 
              Slave_DDL_Groups: 4
Slave_Non_Transactional_Groups: 0
    Slave_Transactional_Groups: 1
1 row in set (0.000 sec)

ERROR: No query specified

 15:42:  [test]>

1032错误的主要原因是主库更新或者是删除的记录在从库上不存在引起的。
处理此种错误一般有两种思路：
1、直接跳过错误执行语句
2、找到错误执行语句，修复从库数据
第一种解决方案会有造成主从不一致的隐患（delete语句可以跳过），第二种是从根本上解决问题比较推荐

/* 本案例不采用这种方式
--  未启用gtid，跳过错误
stop slave;
#表示跳过一步错误，后面的数字可变，（或者N条event，一条一条跳过）
set global sql_slave_skip_counter =1;
start slave;
show slave status/G;

# 还有一种方法跳过所有1032错误
# 更改my.cnf文件，在Replication settings下添加：
#　　slave-skip-errors　＝　1032
# 并重启数据库，然后start salve。
# 注意：因为要重启数据库，不推荐，除非错误事件太多。

*/

# 从库具体操作
# 首先通过查看master的current_pos和当前slave的slave_pos确定需要从哪个gtid_slave_pos继续主从同步
select @@global.gtid_current_pos; # 主库操作
select @@global.gtid_slave_pos;      # 从库操作

 16:15:  [test]> select @@global.gtid_current_pos; # 主库操作
+---------------------------+
| @@global.gtid_current_pos |
+---------------------------+
| 0-330625-75               |
+---------------------------+
1 row in set (0.000 sec)


 16:10:  [test]> select @@global.gtid_slave_pos; # 从库操作
+-------------------------+
| @@global.gtid_slave_pos |
+-------------------------+
| 0-330625-73             |
+-------------------------+
1 row in set (0.000 sec)

/*
然后slave上执行
stop slave; //如果有多个slave复制通道（多主复制），需要全部都停
set global gtid_slave_pos="[new_slave_pos]";
start slave;
show slave status\G;
*/

# 也解析主库对应的binlog，找点。然后向上找到最接近的:SET @@session.gtid_seq_no
#/*!100001 SET @@session.gtid_seq_no=74*//*!*/;
#BEGIN
# /*!*/;
# at 41136
# at 41186
#200101 16:26:10 server id 330625  end_log_pos 41186 CRC32 0x466842d0   Annotate_rows:
#Q> delete from t3 where id <=5
#200101 16:26:10 server id 330625  end_log_pos 41240 CRC32 0x68a74c63   Table_map: `test`.`t3` mapped to number 40
# at 41240
#200101 16:26:10 server id 330625  end_log_pos 41443 CRC32 0x861e9804   Delete_rows: table id 40 flags: STMT_END_F
### DELETE FROM `test`.`t3`
### WHERE
###   @1=1 /* INT meta=0 nullable=0 is_null=0 */
###   @2=0 /* INT meta=0 nullable=0 is_null=0 */
###   @3=20 /* INT meta=0 nullable=0 is_null=0 */

stop slave;
set global gtid_slave_pos=‘0-330625-74‘;
start slave;
show slave status\G;


# 参考资料：https://blog.51cto.com/arkling/2114962

# 一般使用slave_pos，当A->B，A挂掉，B当master，然后A好了，想要做B的slave情况下，
#使用current_pos,因为B以前是主库，没有slave_pos这个值

现在是A挂掉了。B作为新的主库
stop slave ;
reset slave all;

操作：
SET GLOBAL gtid_slave_pos = "";
CHANGE MASTER TO
  MASTER_HOST=‘192.168.142.137‘,
  MASTER_USER=‘repl‘,
  MASTER_PASSWORD=‘replpwd‘,
  MASTER_PORT=3306,
 master_use_gtid=current_pos;
 
 start slave;
 show slave status\G;