Innodb cluster 利用组复制的 pxos 协议,保障数据一致性,组复制支持单主模式和多主模式,html
本文推荐的高可用架构为单主模式下,基于mysqlrouter的 innodb cluster:mysql
系统:centos 7.5linux
Mysql:8.0.12 二进制包sql
Mysqlshell: 8.0.12 rpm 包shell
Mysql router: 8.0.12 二进制包数据库
192.168.181.101 my-manager1 Keepalived、MySQL-shell、MySQL-Router、MySQL-clientbootstrap
192.168.181.102 my-manager2 Keepalived、MySQL-shell、MySQL-Router、MySQL-clientvim
192.168.181.103 mysql-1 MySQL服务端、MySQL-shellcentos
192.168.181.104 mysql-2 MySQL服务端、MySQL-shell服务器
192.168.181.105 mysql-3 MySQL服务端、MySQL-shell
172.16.50.191 uatvfc-db-route01
172.16.50.192 uatvfc-db-route02
172.16.50.193 uatvfc-db-route03
172.16.50.197 uatvfc-db-data01
172.16.50.198 uatvfc-db-data02
172.16.50.199 uatvfc-db-data03
yum -y install gcc glibc libaio
(1)、关闭SElinux
setenforce 0
修改/etc/selinux/config
vim /etc/selinux/config
SELINUX=disabled
(2)、关闭防火墙
systemctl stop firewalld
systemctl disable firewalld
cat>>/etc/sysctl.conf <<EOF
fs.aio-max-nr = 1048576
fs.file-max = 681574400
kernel.shmmax = 137438953472
kernel.shmmni = 4096
kernel.sem = 250 32000 100 200
net.ipv4.ip_local_port_range = 9000 65000
net.core.rmem_default = 262144
net.core.rmem_max = 4194304
net.core.wmem_default = 262144
net.core.wmem_max = 1048586
EOF
cat>>/etc/security/limits.conf <<EOF
mysql soft nproc 65536
mysql hard nproc 65536
mysql soft nofile 65536
mysql hard nofile 65536
EOF
cat>>/etc/pam.d/login <<EOF
session required /lib/security/pam_limits.so
session required pam_limits.so
EOF
cat>>/etc/profile<<EOF
if [ $USER = "mysql" ]; then
ulimit -u 16384 -n 65536
fi
EOF
source /etc/profile
User add mysql
环境变量:
0
cd /data
tar -xzvf mysql-8.0.12-el7-x86_64.tar.gz -C /usr/local
mv /usr/local/mysql-8.0.12-el7-x86_64 /usr/local/mysql
chown -R mysql.mysql /usr/local/mysql
mkdir -p /data/mysql_3310_vfc/{data,log,binlog,conf,tmp}
chown -R mysql.mysql /data/mysql_3310_vfc
mkdir -p /data/mysql_3320_vfc/{data,log,binlog,conf,tmp}
chown -R mysql.mysql /data/mysql_3320_vfc
mkdir -p /data/mysql_3330_vfc/{data,log,binlog,conf,tmp}
chown -R mysql.mysql /data/mysql_3330_vfc
su - mysql
Vim /data/mysql_3310_vfc/conf/my.cnf
[mysqld] lower_case_table_names = 1 user = mysql server_id = 197 port = 3310 default-time-zone = '+08:00' enforce_gtid_consistency = ON gtid_mode = ON binlog_checksum = none default_authentication_plugin = mysql_native_password datadir = /data/mysql_3310_vfc/data pid-file = /data/mysql_3310_vfc/tmp/mysqld.pid socket = /data/mysql_3310_vfc/tmp/mysqld.sock tmpdir = /data/mysql_3310_vfc/tmp/ skip-name-resolve = ON open_files_limit = 65535 # open_files_limit does not take effect table_open_cache = 2000 #################innodb######################## innodb_data_home_dir = /data/mysql_3310_vfc/data innodb_data_file_path = ibdata1:512M;ibdata2:512M:autoextend innodb_buffer_pool_size = 6000M innodb_flush_log_at_trx_commit = 1 innodb_io_capacity = 600 innodb_lock_wait_timeout = 120 innodb_log_buffer_size = 8M innodb_log_file_size = 200M innodb_log_files_in_group = 3 innodb_max_dirty_pages_pct = 85 innodb_read_io_threads = 8 innodb_write_io_threads = 8 innodb_thread_concurrency = 32 innodb_file_per_table innodb_rollback_on_timeout innodb_undo_directory = /data/mysql_3310_vfc/data innodb_log_group_home_dir = /data/mysql_3310_vfc/data |
###################session########################### join_buffer_size = 8M key_buffer_size = 256M bulk_insert_buffer_size = 8M max_heap_table_size = 96M tmp_table_size = 96M read_buffer_size = 8M sort_buffer_size = 2M max_allowed_packet = 64M read_rnd_buffer_size = 32M ############log set################### log-error = /data/mysql_3310_vfc/log/mysqld.err log-bin = /data/mysql_3310_vfc/binlog/binlog log_bin_index = /data/mysql_3310_vfc/binlog/binlog.index max_binlog_size = 500M slow_query_log_file = /data/mysql_3310_vfc/log/slow.log slow_query_log = 1 long_query_time = 10 log_queries_not_using_indexes = ON log_throttle_queries_not_using_indexes = 10 log_slow_admin_statements = ON log_output = FILE,TABLE master_info_file = /data/mysql_3310_vfc/binlog/master.info ##########################mgr set############################## mysqlx_port=33102 mysqlx_socket=/data/mysql_3310_vfc/tmp/mysqlx.sock loose-group_replication_group_name="aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa" loose-group_replication_start_on_boot=off loose-group_replication_local_address= "172.16.50.197:33101" loose-group_replication_group_seeds= "172.16.50.197:33101,172.16.50.198:33101,172.16.50.199:33101" loose-group_replication_bootstrap_group= off loose-group_replication_ip_whitelist="172.16.50.0/24" loose-group_replication_single_primary_mode = on
|
大写的注意: 多主模式存在不少限制和不肯定性问题,不建议使用。具体缘由和读写分离解决方案,看下文 |
多主模式下:
loose-group_replication_single_primary_mode = off
操做流程:业务端链接IP处理 -> GROUP内成员逐个依次主动退出GROUP (所有退出才行)-> 关闭 group_replication_single_primary_mode参数-> 逐个启动GROUP内的SERVER
Set global group_replication_single_primary_mode=off
mysqld --defaults-file=/data/mysql_3310_vfc/conf/my.cnf --initialize --user=mysql 日志中密码:8?_BO,8k.+DX
mysqld --defaults-file=/data/mysql_3320_vfc/conf/my.cnf --initialize --user=mysql 日志中密码:JN&MiMtb5hkj
mysqld --defaults-file=/data/mysql_3330_vfc/conf/my.cnf --initialize --user=mysql 日志中密码:v+-/pcu)M3zo
mysqladmin --defaults-file=my.cnf password 'your password';
mysql -uroot -p -S /data/mysql_3310_vfc/tmp/mysql.sock
mysql -uroot -p -S /data/mysql_3320_vfc/tmp/mysql.sock
mysql -uroot -p -S /data/mysql_3330_vfc/tmp/mysql.sock
alter user root@’localhost’ identified by 'your password';
mysql -uroot -p -S /data/mysql_3306_test/tmp/mysqld.sock
create user root@'127.0.0.1' identified by 'your password';
GRANT all ON *.* TO `root`@`127.0.0.1` WITH GRANT OPTION;
create user root@'172.16.50.197' identified by 'your password';
create user root@'172.16.50.198' identified by 'your password';
create user root@'172.16.50.199' identified by 'your password';
create user root@'172.16.50.191' identified by ‘your password’';
create user root@'172.16.50.192' identified by 'your password’;
create user root@'172.16.50.193' identified by ’your password‘;
关闭主库节点
打包
传递
解压
在作增长从库的复制是,已经对从库作了操做,好比增长帐号等,在最开始操做时可让其不记录日志
SET=0;SQL_LOG_BIN
DDL语句;
DML语句;
SET=1;SQL_LOG_BIN
注意修改 server-id 和删掉 data里面的auto.conf
rpm -ivh mysql-shell-8.0.11-1.el7.x86_64.rpm
rpm -ivh mysql-router*.rpm
路由节点:这里有个小技巧,官方建议将路由节点部署在应用端,这样,应用能够直接使用 app_user@'127.0.0.1‘这样的用户进行访问
使用mysql命令
(1)、检查并配置实例(每一个mysql节点)
mysqlsh --log-level=DEBUG3 日志位置 ~/.mysqlsh/mysqlsh.log
检查实例
dba.configureLocalInstance('root@127.0.0.1:3310');
dba.chekInstanceConfiguration('root@127.0.0.1:3310')
(2)、建立cluster集群(确认每一个SQL节点的实例都完成上述实例配置而且验证成功)
在任意一台mysql实例节点执行如下命令:
mysqlsh
shell.connect('root@172.16.50.197:3310')
var cluster = dba.createCluster('qwCluster1');
若是建立成功输出的信息中会有相似“Cluster successfully created.”的语句
将另外两个节点加入到Cluster集群中
cluster.addInstance('root@172.16.50.198:3310');
cluster.addInstance('root@172.16.50.199:3310');
cluster.addInstance('root@172.16.50.198:3320');
cluster.addInstance('root@172.16.50.199:3320');
cluster.addInstance('root@172.16.50.198:3330');
cluster.addInstance('root@172.16.50.199:3330');
dba.dropMetadataSchema() 清空集群
mysql> stop group_replication;
mysql> reset master; (清空日志,确保和从库的表没有冲突奥,)
mysql> reset slave
mysql> stop group_replication;
mysql> reset master;
mysql> reset slave
[Repl] Slave I/O for channel 'group_replication_recovery': error connecting to master 'mysql_innodb_cluster_r0430970923@mysql3:3306' - retry-time: 60 retries: 1, Error_code: MY-002005
[ERROR] [MY-011582] [Repl] Plugin group_replication reported: 'There was an error when connecting to the donor server. Please check that group_replication_recovery channel credentials and all MEMBER_HOST column values of performance_schema.replication_group_members table are correct and DNS resolvable.'
[ERROR] [MY-011583] [Repl] Plugin group_replication reported: 'For details please check performance_schema.replication_connection_status table and error log messages of Slave I/O for channel group_replication_recovery.'
这个问题郁闷了我好久,
Hostname 是
mysql4
mysql5
mysql6
而我在 /etc/hosts中是
192.168.181.103 mysql-1
192.168.181.103 mysql-2
192.168.181.103 mysql-3
[ERROR] [MY-010586] [Repl] Error running query, slave SQL thread aborted. Fix the problem, and restart the slave SQL thread with "SLAVE START". We stopped at log 'binlog.000007' position 151
[ERROR] [MY-010584] [Repl] Slave SQL for channel 'group_replication_applier': Error executing row event: 'Unknown database 'mysql_innodb_cluster_metadata'', Error_code: MY-001049
重建master reset master
组复制建议,事物隔离级别,read commit
group_replication_enforce_update_everywhere_checks=ON
group_replication_enforce_update_everywhere_checks=ON
多主模式下不支持。这部份内容主要涉及到几个系统表格,有点相似于 SQL SERVER中的DMV视图,详见下表。
库名 |
表名 |
type |
description |
performance_schema |
replication_group_members |
重要,经常使用 |
查看GROUP成员。 |
performance_schema |
replication_group_member_stats |
重要,经常使用 |
当前SERVER在GROUP中的同步状况,查看applier通道的同步状况。 |
performance_schema |
replication_connection_stats |
重要,经常使用 |
当前server中各个通道的使用状况,applier通道是必定有显示,recovery通道看是否使用过,若是有则显示,没有则不显示。 |
performance_schema |
replication_applier_stats |
重要,经常使用 |
当前server中各个通道是否启用。 |
performance_schema |
global_status |
重要,经常使用 |
单主模式下,能够查看当前主库是哪一个。 |
performance_schema |
replication_applier_configuration |
不经常使用,了解便可 |
|
performance_schema |
replication_applier_status_by_coordinator |
不经常使用,了解便可 |
|
performance_schema |
replication_applier_status_by_worker |
不经常使用,了解便可 |
|
performance_schema |
replication_connection_configuration |
不经常使用,了解便可 |
|
Mysql |
slave_master_info |
重要,不经常使用 |
设置了master_info_repository=TABLE,因此master的相关信息会存储在这个表格。 |
Mysql |
slave_relay_log_info |
重要,不经常使用 |
设置了relay_log_info_repository=TABLE,因此master的相关信息会存储在这个表格。 |
l 组复制成员
select * from performance_schema.replication_group_members;
l 组复制状态
select * from performance_schema.global_status where variable_name like '%group%';
l 单主模式下主库是哪一个
SELECT * FROM performance_schema.replication_group_members;
SELECT * FROM performance_schema. global_status;
l 检查数据库是否正常提供读写服务
show global variables like 'super%';
SELECT * FROM performance_schema.replication_group_members;
若是super_read_only是启动的,那么该成员仅提供读服务;
若是super_read_only是关闭的,而且 replication_group_members 中正常的成员n 知足 2n+1 > 整个GROUP成员个数,而且该成员的 member state是online,则该成员可提供读写服务。
l 检查数据库是否复制出现问题
能够经过表格replication_group_members ,replication_group_member_stats ,replication_connection_stats ,replication_applier_stats 查看
重点注意各个 组成员的 ERROR LOG详细信息,由于报错描述最清楚都在这里了。
# 在启动第一个节点时
SET GLOBAL group_replication_bootstrap_group=ON; #表示本节点做为组复制的起始节点
START GROUP_REPLICATION; #开启组复制
SET GLOBAL group_replication_bootstrap_group=OFF; # 关闭组复制做为起始节点
# 启动其余节点
START GROUP_REPLICATION;
#关闭组复制
stop group replication
mysqlsh --log-level=debug3
The following functions are currently supported.
- checkInstanceConfiguration 校验实例配置
- configureInstance 配置实例集群
- configureLocalInstance 配置本地实例 8.0.11之后放弃了
- createCluster 建立集群.
- dropMetadataSchema 删除集群.
- getCluster 提取cluster 从元数据中
- rebootClusterFromCompleteOutage 当集群成员都OFFLINE 能够启动起来.
e.g. dba.help('configureInstance')
shell.connect('root@192.168.181.103:3306')
var cluster = dba.getCluster()
cluster.status() #集群状态查看
var cluster = dba.createCluster('qwCluster')
cluster.describe();
使用一个存在的组复制建立集群
var cluster = dba.createCluster('qwCluster', {adoptFromGR: true});
dba.dropMetadataSchema()
cluster.removeInstance('root@localhost:3310')
cluster.rejoinInstance()
cluster.checkInstanceState('root@127.0.0.1:3306')
dba.configureInstance('root@192.168.181.103:3306',{mycnfPath:'/data/mysql_3306_test/conf/my.cnf',clusterAdmin:'cadmin@mysql-1%',clusterAdminPassword:'fangfang'})
var cluster = dba.rebootClusterFromCompleteOutage();
官网说,若是配置了innodbcluster 请物本身配置 集群信息
mysqlrouter --bootstrap root@mysql3:3306 --user=mysqlrouter
设置开机启动
Systemctl enable mysqlrouter
Systemctl start mysqlrouter
在远端访问路由执行命令
mysql -uroot -p -h192.168.181.101 -P6446 -e "select @@hostname";
能够看到访问的路径是轮训的
l (单主模式)
l (多主模式)
将全部节点所有关机,重启后
SET GLOBAL group_replication_bootstrap_group=ON;
START GROUP_REPLICATION;
SET GLOBAL group_replication_bootstrap_group=OFF;
全部节点所有关闭的状况下,宕机状况下:
var cluster = dba.rebootClusterFromCompleteOutage(‘qwCluster1’);
启动一个节点,而后其余节点重启自动加入
mysqlsh --log-level=debug3
主节点关闭,会当即切换一个从节点为主库,mysqlroute 无感知
启动新的节点,会自动恢复复制状态,无需关心。
从节点关闭,不影响读写,从新启动数据库后自动恢复同步。
等同于,有一个节点已经挂掉,没法启动。
此时须要添加节点
mysqlbackup是一个热备份工具、也就是说它不像mysqldump那样给表上一个全局锁,因为mysqldump上了这个锁,因此就形成客户端只能对数据库进行读操做不能写,这也就是称mysqldump为温备份的缘由。可是mysqlbackup真的有这么吊吗?答案是并无。对于innodb引擎的表mysqlbackup 热备的;可是对于非innodb表mysqlbackup就只能温备了,缘由是这类引擎不支持事务也就是说不能经过事务日志来保证备份的一致性,因此就只能给表加上一个全局锁来解决了。为了获得一致的备份mysqlbackup要不停的去追踪mysql数据库的sln号,也就是说mysqlbackup要执行备份那么它必定要链接上数据库。mysqlbackup对数据库的备份是经过复制文件的方式进行的,也就是说mysqlbackup要和数据库在同一台机器上。
Mysqlbackup 是企业版mysql中的,但能够拿来使用,登陆oracle云下载,搜索mysql backup,下载二进制文件
mysql-commercial-backup-8.0.11-linux-glibc2.12-x86_64.tar.gz
传到主机上并解压,
会有 bin 和lib 文件夹,将其中的文件拷贝到 /usr/local/mysql 对应的目录中, 这里要把 软链接,库所有拷贝的,否则会报错
(官方指导命令)
CREATE USER 'mysqlbackup'@'localhost' IDENTIFIED BY 'new-password';
GRANT RELOAD ON *.* TO 'mysqlbackup'@'localhost';
GRANT CREATE, INSERT, DROP, UPDATE ON mysql.backup_progress TO 'mysqlbackup'@'localhost';
GRANT CREATE, INSERT, SELECT, DROP, UPDATE ON mysql.backup_history TO 'mysqlbackup'@'localhost';
GRANT REPLICATION CLIENT ON *.* TO 'mysqlbackup'@'localhost';
GRANT SUPER ON *.* TO 'mysqlbackup'@'localhost';
GRANT PROCESS ON *.* TO 'mysqlbackup'@'localhost';
GRANT ALTER ON mysql.backup_history TO 'mysqlbackup'@'localhost';
GRANT LOCK TABLES, SELECT, CREATE, DROP, FILE ON *.* TO 'mysqlbackup'@'localhost';
GRANT CREATE, INSERT, DROP, UPDATE ON mysql.backup_sbt_history TO 'mysqlbackup'@'localhost';
mysqlbackup --user=mybackup --password=131417 --backup-image=/data/backup/mysql_server.mbi --backup-dir=/data/backup/temp backup-to-image –with-timestap
mysqlbackup –backup-dir=/data/backup/tmp –socket=/data/mysql_3306_test/tmp/mysql.sock –user=root –password=NTg1Z@mYxZjdhZWI –backup-image=/data/backup/full_20180718.bki --with-timestap backup-to-image
mysqlbackup --backup-dir=/backups --backup-image=- backup-to-image > /backup/mybackup.mbi
--user:用户名。
--password:密码。
--port:端口,默认值为3306。
--backup-dir:能够当作是mysqlback的工做目录,临时用的。
--backup-image:备份文件名。
backup-to-image:把全部的备份信息输出到一个备份文件当中
--with-timepstap: 会自动生成一个带日期文件夹
备份的操做:先进行全备后进行增备
mysqlbackup --user=mysqlbackup --password=123 --backup-image=backup.mbi --backup-dir=/backup --with-timestamp backup-to-image #先备份成image
mysqlbackup --defaults-file=/etc/my.cnf --user=mysqlbackup --password=123 --with-timestamp --incremental --incremental-backup-dir=/backup/ --incremental-base=dir:/backup/2017-09-18_15-53-57 backup #在原来备份的基础上进行增量备份
#########备份到其余服务器
The following command streams the backup as a single-file output to a remote host to be saved under
the file name my_backup.img (--backup-dir=/tmp designates the directory for storing temporary files
rather than the final output file):
mysqlbackup --defaults-file=~/my_backup.cnf --backup-image=- --backup-dir=/tmp backup-to-image | \
ssh <user name>@<remote host name> 'cat > ~/backups/my_backup.img'
./mysqlbackup --backup-image=/home/admin/backups/my.mbi validate
mysqlbackup --backup-image=/backup/my.mbi list-image
mysqlbackup --backup-dir=/backup/backup --backup-image=/backup/2017-09-18_15-53-57/backup.mbi image-to-backup-dir
mysqlbackup --defaults-file=/backup/2017-09-15_17-06-07/server-my.cnf --datadir=/data/mysql --backup-dir=/backup/2017-09-15_17-06-07 copy-back
1 主库作全备份
备份会锁myisam表,如master是生产库,须要注意,可留意业务进程是否有活动的。
/test/mysql/base/bin/mysqlbackup --defaults-file=/test/my3352/my.cnf --socket=/test/my3352/var/mysql.sock --user=root --password --with-timestamp --backup-dir=/test/my3352/backup backup-and-apply-log
2 备份文件拷到从库主机,作恢复
保证从库的data和log目录为空,用从库的my.cnf进行恢复
恢复时可能也会将主库备份的binlog和my.cnf直接恢复到data目录下,能够清理掉
mv data data_bak
mv log log_bak
mkdir data
mkdir log
停从库,清空目录
mysqlbackup --defaults-file=/test/my3309/my.cnf --backup-dir=/test/myback/BOT/temp copy-back
$ ps -ef | grep my3330
3 启动从库
Cd $MYSQL_HOME
mysqld_safe --defaults-file=/test/mydat/$PORT/my.cnf &
mysql -uroot --socket=/test/my3330/var/mysql.sock
4 执行备份文件meta目录下backup_gtid_executed.sql文件的SET @@GLOBAL.GTID_PURGED语句(可source backup_gtid_executed.sql)
mysql> start slave ;
mysql> show slave status\G
5 按标准change master to
mysql> change master to master_host='10.33.45.152', master_port=3330,master_user='dbsync', master_password='********', master_auto_position=1;
6 start slave。
可能会报ERROR 1872,能够reset slave后,再从新change master.
ERROR 1872 (HY000): Slave failed to initialize relay log info structure from the repository
将备份恢复到其余主机上去