18 Star 92 Fork 50

万里数据库 / GreatSQL-Doc

Create your Gitee Account
Explore and code with more than 12 million developers,Free private repositories !:)
Sign up
Clone or Download
using-greatsql-to-build-mgr-and-node-manage.md 27.52 KB
Copy Edit Raw Blame History
GreatSQL authored 2023-12-28 14:43 . 内容微调,更新标题等

一文掌握GreatSQL MGR集群的部署和运维

[toc]

本文详细介绍如何在单机环境下,利用GreatSQL构建一个3节点的MGR集群,并用mysqld_multi进行管理( 也建议采用systemd来管理GreatSQL多实例服务,参考:利用systemd管理MySQL单机多实例 )。

为了简单起见,这个MGR集群采用单主(single-primary)模式,不采用多主(multi-primary)模式。

构建完MGR集群后,再添加一个新节点,以及模拟进行滚动升级和切主等其他操作。

除了MySQL官方社区版本外,如果想体验更可靠、稳定、高效的MGR,推荐使用GreatSQL版本。本文采用GreatSQL 8.0.22版本,关于这个版本的说明详见 GreatSQL,打造更好的MGR生态

P.S,单机模式下,如果要部署多实例并构建MGR集群,要注意避免TCP self-connect的问题,详见 bug#98151,如果采用GreatSQL版本就没这个问题了。

1. 运行环境

GreatSQL二进制包放在 /usr/local/ 下,即 basedir = /usr/local/GreatSQL-8.0.22

三个实例按下面规划分配:

实例 端口 datadir
GreatSQL-01 3306 /data/GreatSQL/mgr01/
GreatSQL-02 3307 /data/GreatSQL/mgr02/
GreatSQL-03 3308 /data/GreatSQL/mgr03/

输入图片说明

2、准备my.cnf配置文件

[mysqld]
basedir=/usr/local/GreatSQL-8.0.22
log_timestamps=SYSTEM
user = mysql
log_error_verbosity = 3

log-bin=binlog
binlog-format=row
log_slave_updates=ON
binlog_checksum=CRC32

master-info-repository=TABLE
relay-log-info-repository=TABLE
gtid-mode=on
enforce-gtid-consistency=true
binlog_transaction_dependency_tracking=writeset
transaction_write_set_extraction=XXHASH64
slave_parallel_type = LOGICAL_CLOCK
slave_parallel_workers=128 #可以设置为逻辑CPU数量的2-4倍
sql_require_primary_key=1
slave_preserve_commit_order=1
slave_checkpoint_period=2

#mgr
loose-plugin_load_add='mysql_clone.so'
loose-plugin_load_add='group_replication.so'

#所有节点的group_replication_group_name值必须相同
#这是一个标准的UUID格式,可以手动指定,也可以用随机生成的UUID
loose-group_replication_group_name="aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaa1"

#指定MGR集群各节点的IP+端口,这个端口是专用于MGR的,不是平常所说的mysqld实例端口
#如果是在多节点上部署MGR集群时,要注意这个端口是否会被防火墙拦截
loose-group_replication_group_seeds= "127.0.0.1:33061,127.0.0.1:33071,127.0.0.1:33081"

#不建议启动mysqld的同时也启动MGR服务
loose-group_replication_start_on_boot=off

#默认不要作为MGR集群引导节点,有需要时再手动执行并立即改回OFF状态
loose-group_replication_bootstrap_group=off

#当退出MGR后,把该实例设置为read_only,避免误操作写入数据
loose-group_replication_exit_state_action=READ_ONLY

#一般没什么必要开启流控机制
loose-group_replication_flow_control_mode = "DISABLED"

#【强烈】建议只用单主模式,如果是实验目的,可以尝试玩玩多主模式
loose-group_replication_single_primary_mode=ON

[mysqld_multi]
mysqld = /usr/local/GreatSQL-8.0.22/bin/mysqld
log = /data/GreatSQL/mysqld_multi.log
mysqladmin = /usr/local/GreatSQL-8.0.22/bin/mysqladmin
user=root

[mysqld3306]
datadir=/data/GreatSQL/mgr01
socket=/data/GreatSQL/mgr01/mysql.sock
port=3306
server_id=3306
log-error=/data/GreatSQL/mgr01/error.log
#指定本节点的IP+端口
loose-group_replication_local_address= "127.0.0.1:33061"

[mysqld3307]
datadir=/data/GreatSQL/mgr02
socket=/data/GreatSQL/mgr02/mysql.sock
port=3307
server_id=3307
log-error=/data/GreatSQL/mgr02/error.log
loose-group_replication_local_address= "127.0.0.1:33071"

[mysqld3308]
datadir=/data/GreatSQL/mgr03
socket=/data/GreatSQL/mgr03/mysql.sock
port=3308
server_id=3308
log-error=/data/GreatSQL/mgr03/error.log
loose-group_replication_local_address= "127.0.0.1:33081"

在这份配置文件中,[mysqld] 这部分内容是所有实例都会读取应用的,而在 [mysqld3306] 这部分配置,才是 3306 端口这个实例所独有的。

在构建MGR集群中,要保证集群各节点的 group_replication_group_name 选项值一样才行,否则就是不同的集群了。

另外,如果有防火墙的话,注意要开放各端口间的访问规则,否则无法启动MGR。

3、初始化MySQL实例

先手动创建相应的datadir,并修改目录属主为mysql用户:

[root@greatsql]# mkdir -p /data/GreatSQL/{mgr01,mgr02,mgr03}
[root@greatsql]# chown -R mysql.mysql /data/GreatSQL

执行下面的命令进行MySQL实例初始化,会自动创建InnoDB系统表空间、redo log、undo log的文件:

[root@greatsql]# /usr/local/GreatSQL-8.0.22/bin/mysqld --no-defaults --datadir=/data/GreatSQL/mgr01 --initialize --user=mysql

[System] [MY-013169] [Server] /usr/local/GreatSQL-8.0.22/bin/mysqld (mysqld 8.0.22-13) initializing of server in progress as process 18688
[System] [MY-013576] [InnoDB] InnoDB initialization has started.
[System] [MY-013577] [InnoDB] InnoDB initialization has ended.
[Note] [MY-010454] [Server] A temporary password is generated for root@localhost: h<GL%Lr:v66W

可以看到,在输出的日志中打印了root账号的临时密码(最后一行),启动mysqld实例后,首次用这个密码登入,就要立即进行修改,否则其他什么也做不了:

用同样的方法,也分别完成mgr02、mgr03的初始化。

接下来,分别启动三个mysqld实例:

[root@greatsql]# /usr/local/GreatSQL-8.0.22/bin/mysqld_multi start 3306
[root@greatsql]# /usr/local/GreatSQL-8.0.22/bin/mysqld_multi start 3307
[root@greatsql]# /usr/local/GreatSQL-8.0.22/bin/mysqld_multi start 3308

能看到datadir下的文件目录大概像这样的:

-rw-r----- 1 mysql mysql       56 Jun  4 10:44 auto.cnf
-rw-r----- 1 mysql mysql      401 Jun  4 10:46 binlog.000001
-rw-r----- 1 mysql mysql       16 Jun  4 10:46 binlog.index
-rw------- 1 mysql mysql     1676 Jun  4 10:44 ca-key.pem
-rw-r--r-- 1 mysql mysql     1120 Jun  4 10:44 ca.pem
-rw-r--r-- 1 mysql mysql     1120 Jun  4 10:44 client-cert.pem
-rw------- 1 mysql mysql     1680 Jun  4 10:44 client-key.pem
-rw-r----- 1 mysql mysql     8800 Jun  4 10:46 error.log
-rw-r----- 1 mysql mysql   196608 Jun  4 10:48 #ib_16384_0.dblwr
-rw-r----- 1 mysql mysql  8585216 Jun  4 10:44 #ib_16384_1.dblwr
-rw-r----- 1 mysql mysql     6274 Jun  4 10:44 ib_buffer_pool
-rw-r----- 1 mysql mysql 12582912 Jun  4 10:46 ibdata1
-rw-r----- 1 mysql mysql 50331648 Jun  4 10:48 ib_logfile0
-rw-r----- 1 mysql mysql 50331648 Jun  4 10:44 ib_logfile1
-rw-r----- 1 mysql mysql 12582912 Jun  4 10:46 ibtmp1
drwxr-x--- 2 mysql mysql      168 Jun  4 10:46 #innodb_temp
drwxr-x--- 2 mysql mysql      143 Jun  4 10:44 mysql
-rw-r----- 1 mysql mysql 25165824 Jun  4 10:46 mysql.ibd
srwxrwxrwx 1 mysql mysql        0 Jun  4 10:46 mysql.sock
-rw------- 1 mysql mysql        6 Jun  4 10:46 mysql.sock.lock
drwxr-x--- 2 mysql mysql     8192 Jun  4 10:46 performance_schema
-rw------- 1 mysql mysql     1680 Jun  4 10:44 private_key.pem
-rw-r--r-- 1 mysql mysql      452 Jun  4 10:44 public_key.pem
-rw-r--r-- 1 mysql mysql     1120 Jun  4 10:44 server-cert.pem
-rw------- 1 mysql mysql     1676 Jun  4 10:44 server-key.pem
drwxr-x--- 2 mysql mysql       28 Jun  4 10:44 sys
-rw-r----- 1 mysql mysql 10485760 Jun  4 10:48 undo_001
-rw-r----- 1 mysql mysql 10485760 Jun  4 10:46 undo_002
-rw-r----- 1 mysql mysql        6 Jun  4 10:46 greatsql.pid

4、构建MGR集群

4.1 构建MGR集群前的准备工作

因为在配置文件中已经指定了要加载 group_replicationmysql_clone 两个 plugin,如无意外,应该都已经加载成功:

[root@GreatSQL][(3306)]> show plugins;
+---------------------------------+----------+--------------------+----------------------+---------+
| Name                            | Status   | Type               | Library              | License |
+---------------------------------+----------+--------------------+----------------------+---------+
| binlog                          | ACTIVE   | STORAGE ENGINE     | NULL                 | GPL     |
...
| clone                           | ACTIVE   | CLONE              | mysql_clone.so       | GPL     |
| group_replication               | ACTIVE   | GROUP REPLICATION  | group_replication.so | GPL     |
+---------------------------------+----------+--------------------+----------------------+---------+

看到确实已经加载了。

如果没有被正确加载,就需要查看日志文件确认什么原因无法加载。

也可以尝试手动加载这两个plugin:

[root@GreatSQL][(3306)]> INSTALL PLUGIN group_replication SONAME 'group_replication.so';
[root@GreatSQL][(3306)]> INSTALL PLUGIN clone SONAME 'mysql_clone.so';

clone plugin 的作用后面再介绍。

4.2 配置MGR集群PRIMARY节点

接下来创建MGR所需要的账户,并授权:

[root@GreatSQL][(3306)]> CREATE USER repl@'%' IDENTIFIED WITH MYSQL_NATIVE_PASSWORD BY 'repl';
[root@GreatSQL][(3306)]> GRANT REPLICATION SLAVE, BACKUP_ADMIN ON *.* TO `repl`@`%`;

因为是一个刚初始化的干净系统,也为了简单起见,执行下面的命令再重置一下:

[root@GreatSQL][(3306)]> reset master; reset slave all;

【提醒】生产环境中请勿这么做。后面会有其他文章介绍如何对已上线的MGR集群再加入新的节点。

创建MGR复制通道:

[root@GreatSQL][(3306)]> CHANGE MASTER TO MASTER_USER='repl', MASTER_PASSWORD='repl' FOR CHANNEL 'group_replication_recovery';

上述所有操作,在其他几个节点上都重复执行一遍。

接着,【重要的一步】来了,登入被选中作为 PRIMARY 节点的mgr01实例,执行下面的命令:

[root@GreatSQL][(3306)]> set global group_replication_bootstrap_group=ON;

这个命令的作用是在MGR集群的PRIMARY节点第一次被启动时,用于引导MGR集群的。

在其他节点启动时,记住【不要】将选项 group_replication_bootstrap_group 设置为ON,否则会独立拉起一个新的MGR集群。

之后就可以在MGR集群的 PRIMARY 节点上启动组复制线程了:

[root@GreatSQL][(3306)]> start group_replication;
Query OK, 0 rows affected (2.14 sec)

[root@GreatSQL][(3306)]> select * from performance_schema.replication_group_members;
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| CHANNEL_NAME              | MEMBER_ID                            |  MEMBER_HOST | MEMBER_PORT | MEMBER_STATE | MEMBER_ROLE | MEMBER_VERSION |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| group_replication_applier | 0fbb2cfd-c4d9-11eb-8747-525400e2078a | greatsql-mgr |        3306 | ONLINE       | PRIMARY     | 8.0.22         |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+

此时可以看到MGR集群已启动,且只有PRIMARY节点。

集群启动后,记得立即将该选项重置为 OFF。

[root@GreatSQL][(3306)]> set global group_replication_bootstrap_group=OFF;

接下来在mgr02、mgr03节点上也执行 start group_replication 启动MGR服务,【记得要设置 】group_replication_bootstrap_group=OFF

4.3 查看MGR集群状态

所有实例都启动MGR服务后,再次查看集群状态:

[root@GreatSQL][(3308)]> select * from performance_schema.replication_group_members;
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| CHANNEL_NAME              | MEMBER_ID                            |  MEMBER_HOST | MEMBER_PORT | MEMBER_STATE | MEMBER_ROLE | MEMBER_VERSION |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| group_replication_applier | 0fbb2cfd-c4d9-11eb-8747-525400e2078a | greatsql-mgr |        3306 | ONLINE       | PRIMARY     | 8.0.22         |
| group_replication_applier | 1778e87a-c4d9-11eb-bf33-525400e2078a | greatsql-mgr |        3307 | ONLINE       | SECONDARY   | 8.0.22         |
| group_replication_applier | 1c8ebfa6-c4d9-11eb-80d1-525400e2078a | greatsql-mgr |        3308 | ONLINE       | SECONDARY   | 8.0.22         |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+

因为是在同一个主机上启动多实例构建的MGR,所以 MEMBER_HOST 的值是一样的。

一个在单机上由三个实例组成的MGR集群构建完毕。

4.4 测试数据读写

在MGR集群运行过程中,只有PRIMARY节点上允许同时读写数据,其他节点只能只读数据,不能写入

在PRIMARY节点上,创建新的库表并写入几行数据:

[root@GreatSQL][(3306)]> create database greatsql;
[root@GreatSQL][(3306)]> use greatsql;
[root@GreatSQL][(3306)][greatsql]> create table t1(id int primary key);
[root@GreatSQL][(3306)][greatsql]> insert into t1 values (rand()*1024), (rand()*1024), (rand()*1024);
Query OK, 3 rows affected (0.01 sec)
Records: 3  Duplicates: 0  Warnings: 0

[root@GreatSQL][(3306)][greatsql]> select * from t1;
+-----+
| id  |
+-----+
| 105 |
| 423 |
| 557 |
+-----+
3 rows in set (0.00 sec)

在另外两个节点上查看数据:

[root@GreatSQL][(3308)]> select * from greatsql.t1;
+-----+
| id  |
+-----+
| 105 |
| 423 |
| 557 |
+-----+
3 rows in set (0.00 sec)

可以读取到刚新写入的数据。

再次查看MGR的applier线程工作状态:

[root@GreatSQL][(3306)]> select * from performance_schema.replication_connection_status where channel_name = 'group_replication_applier'\G
*************************** 1. row ***************************
                                      CHANNEL_NAME: group_replication_applier
                                        GROUP_NAME: aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaa1
                                       SOURCE_UUID: aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaa1
                                         THREAD_ID: NULL
                                     SERVICE_STATE: ON   <---状态为ON,正常
                         COUNT_RECEIVED_HEARTBEATS: 0
                          LAST_HEARTBEAT_TIMESTAMP: 0000-00-00 00:00:00.000000
                          RECEIVED_TRANSACTION_SET: aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaa1:1-19  <--- GTID在持续变化
                                 LAST_ERROR_NUMBER: 0
                                LAST_ERROR_MESSAGE:
                              LAST_ERROR_TIMESTAMP: 0000-00-00 00:00:00.000000
                           LAST_QUEUED_TRANSACTION: aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaa1:19
 LAST_QUEUED_TRANSACTION_ORIGINAL_COMMIT_TIMESTAMP: 0000-00-00 00:00:00.000000
LAST_QUEUED_TRANSACTION_IMMEDIATE_COMMIT_TIMESTAMP: 0000-00-00 00:00:00.000000
     LAST_QUEUED_TRANSACTION_START_QUEUE_TIMESTAMP: 2021-06-04 15:53:55.395605
       LAST_QUEUED_TRANSACTION_END_QUEUE_TIMESTAMP: 2021-06-04 15:53:55.395630
                              QUEUEING_TRANSACTION:
    QUEUEING_TRANSACTION_ORIGINAL_COMMIT_TIMESTAMP: 0000-00-00 00:00:00.000000
   QUEUEING_TRANSACTION_IMMEDIATE_COMMIT_TIMESTAMP: 0000-00-00 00:00:00.000000
        QUEUEING_TRANSACTION_START_QUEUE_TIMESTAMP: 0000-00-00 00:00:00.000000

以及查看复制组各成员的状态:

[root@GreatSQL][(3306)]> select * from performance_schema.replication_group_member_stats\G
*************************** 1. row ***************************
                              CHANNEL_NAME: group_replication_applier
                                   VIEW_ID: 16227931944218245:4
                                 MEMBER_ID: 0fbb2cfd-c4d9-11eb-8747-525400e2078a
               COUNT_TRANSACTIONS_IN_QUEUE: 0   <--- 等待冲突检测的事务队列大小
                COUNT_TRANSACTIONS_CHECKED: 0
                  COUNT_CONFLICTS_DETECTED: 0
        COUNT_TRANSACTIONS_ROWS_VALIDATING: 0
        TRANSACTIONS_COMMITTED_ALL_MEMBERS: aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaa1:1-19
            LAST_CONFLICT_FREE_TRANSACTION:
COUNT_TRANSACTIONS_REMOTE_IN_APPLIER_QUEUE: 0   <--- 本地等待apply的事务队列大小
         COUNT_TRANSACTIONS_REMOTE_APPLIED: 3
         COUNT_TRANSACTIONS_LOCAL_PROPOSED: 0
         COUNT_TRANSACTIONS_LOCAL_ROLLBACK: 0
         ...

5、进一步,再增加一个新节点

如果想对MGR集群扩展读性能,可以增加新的SECONDARY节点。

按照前面的方法,先初始化一个新的实例 mgr04,它运行的端口是 3309。然后利用 clone plugin(主从节点都必须要启用clone plugin) 从现有其他节点复制数据过来,再加入MGR集群。

运行clone复制数据需要 BACKUP_ADMIN 权限(复制源、目标两个节点都需要),前面已经授予了。

首先,设置clone donor节点:

# 尽量从SECONDARY节点复制数据,不从PRIMARY节点复制
[root@GreatSQL][(3309)]> set global clone_valid_donor_list='127.0.0.1:3307';

开始复制数据:

[root@GreatSQL][(3309)]> clone instance from repl@127.0.0.1:3307 identified by 'repl';

#clone结束后,会自动重启mysqld实例
#但因为该实例没有用systemd服务管理起来,所以需要手动启动进程
ERROR 3707 (HY000): Restart server failed (mysqld is not managed by supervisor process).

再次启动3309端口实例,登入查询,就能看到从其他节点复制过来的数据了:

[root@GreatSQL][(3309)]> select * from greatsql.t1;
+-----+
| id  |
+-----+
| 105 |
| 423 |
| 557 |
+-----+
3 rows in set (0.00 sec)

在mgr04实例上启动MGR服务:

[root@GreatSQL][(3309)]> start group_replication;
Query OK, 0 rows affected (2.85 sec)

[root@GreatSQL][(3309)]> select * from performance_schema.replication_group_members;
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| CHANNEL_NAME              | MEMBER_ID                            |  MEMBER_HOST | MEMBER_PORT | MEMBER_STATE | MEMBER_ROLE | MEMBER_VERSION |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| group_replication_applier | 0fbb2cfd-c4d9-11eb-8747-525400e2078a | greatsql-mgr |        3306 | ONLINE       | PRIMARY     | 8.0.22         |
| group_replication_applier | 1778e87a-c4d9-11eb-bf33-525400e2078a | greatsql-mgr |        3307 | ONLINE       | SECONDARY   | 8.0.22         |
| group_replication_applier | 1c8ebfa6-c4d9-11eb-80d1-525400e2078a | greatsql-mgr |        3308 | ONLINE       | SECONDARY   | 8.0.22         |
| group_replication_applier | bc037d6c-c4de-11eb-a5b8-525400e2078a | greatsql-mgr |        3309 | ONLINE       | SECONDARY   | 8.0.22         |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+

可以看到,新节点加入成功了。

输入图片说明

6、再进一步,切换主节点

如果需要升级MGR集群中各节点的MySQL版本,则可以进行滚动升级

先升级完全部SECONDARY节点,再将PRIMARY节点停掉下线,最后升级,MGR集群会自动选择其他已升级完的节点作为新的PRIMARY节点,等到原来的PRIMARY节点也升级完后再加入 回来即可,这就完成整个集群所有节点的升级工作了。

6.1 先升级SECONDARY节点

现在要升级mgr4节点的MySQL版本,需要先停掉MGR服务:

[root@GreatSQL][(3309)]> stop group_replication;

再停掉mysqld进程后,在my.cnf中增加一行配置:

upgrade=AUTO

替换/指定新的MySQL二进制程序文件(相同大版本,只有相近几个小版本的差异时可以这么做),再次启动mysqld进程,即可实现自动升级。

这是MySQL 8.0.16之后的升级新方法,在8.0.16之前,需要手动执行 mysql_upgrade 二进制程序进行升级。

启动过程中,能看到类似下面的日志:

[System] [MY-013381] [Server] Server upgrade from '80022' to '80025' started.
[Note] [MY-013386] [Server] Running queries to upgrade MySQL server.
[Note] [MY-013387] [Server] Upgrading system table data.
[Note] [MY-013385] [Server] Upgrading the sys schema.

之后再启动MGR服务,就能看到各节点的MySQL版本不同了。

[root@GreatSQL][(3309)]> select * from performance_schema.replication_group_members;
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| CHANNEL_NAME              | MEMBER_ID                            |  MEMBER_HOST | MEMBER_PORT | MEMBER_STATE | MEMBER_ROLE | MEMBER_VERSION |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| group_replication_applier | 0fbb2cfd-c4d9-11eb-8747-525400e2078a | greatsql-mgr |        3306 | ONLINE       | PRIMARY     | 8.0.22         |
| group_replication_applier | 1778e87a-c4d9-11eb-bf33-525400e2078a | greatsql-mgr |        3307 | ONLINE       | SECONDARY   | 8.0.25         |
| group_replication_applier | 1c8ebfa6-c4d9-11eb-80d1-525400e2078a | greatsql-mgr |        3308 | ONLINE       | SECONDARY   | 8.0.25         |
| group_replication_applier | bc037d6c-c4de-11eb-a5b8-525400e2078a | greatsql-mgr |        3309 | ONLINE       | SECONDARY   | 8.0.25         |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+

如上图所示,只剩下mgr01节点还没升级。

从MySQL 8.0.16开始,新增MGR协议要求,比如使用相同的通信协议版本才能组成MGR集群,而8.0.22和8.0.25是相同的,所以可以在同一个集群里。

#在mgr01上执行
[root@GreatSQL][(3306)]> select version(), group_replication_get_communication_protocol();
+-----------+------------------------------------------------+
| version() | group_replication_get_communication_protocol() |
+-----------+------------------------------------------------+
| 8.0.22-13 | 8.0.16                                         |
+-----------+------------------------------------------------+

#在mgr04上执行
[root@GreatSQL][(3309)]> select version(), group_replication_get_communication_protocol();
+-----------+------------------------------------------------+
| version() | group_replication_get_communication_protocol() |
+-----------+------------------------------------------------+
| 8.0.25    | 8.0.16                                         |
+-----------+------------------------------------------------+

6.2 再升级PRIMARY节点

现在关闭mgr01节点后,剩下的三个节点会完成自动选主:

+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| CHANNEL_NAME              | MEMBER_ID                            |  MEMBER_HOST | MEMBER_PORT | MEMBER_STATE | MEMBER_ROLE | MEMBER_VERSION |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| group_replication_applier | 1778e87a-c4d9-11eb-bf33-525400e2078a | greatsql-mgr |        3307 | ONLINE       | PRIMARY     | 8.0.25         |
| group_replication_applier | 1c8ebfa6-c4d9-11eb-80d1-525400e2078a | greatsql-mgr |        3308 | ONLINE       | SECONDARY   | 8.0.25         |
| group_replication_applier | bc037d6c-c4de-11eb-a5b8-525400e2078a | greatsql-mgr |        3309 | ONLINE       | SECONDARY   | 8.0.25         |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+

自动选择了 mgr02 节点作为新的PRIMARY节点(未设定各节点权重值时,则按照 MEMBER_ID 的顺序依次选主)。

待到mgr01节点也完成升级,重新加回集群后:

+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| CHANNEL_NAME              | MEMBER_ID                            |  MEMBER_HOST | MEMBER_PORT | MEMBER_STATE | MEMBER_ROLE | MEMBER_VERSION |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| group_replication_applier | 0fbb2cfd-c4d9-11eb-8747-525400e2078a | greatsql-mgr |        3306 | ONLINE       | SECONDARY   | 8.0.25         |
| group_replication_applier | 1778e87a-c4d9-11eb-bf33-525400e2078a | greatsql-mgr |        3307 | ONLINE       | PRIMARY     | 8.0.25         |
| group_replication_applier | 1c8ebfa6-c4d9-11eb-80d1-525400e2078a | greatsql-mgr |        3308 | ONLINE       | SECONDARY   | 8.0.25         |
| group_replication_applier | bc037d6c-c4de-11eb-a5b8-525400e2078a | greatsql-mgr |        3309 | ONLINE       | SECONDARY   | 8.0.25         |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+

输入图片说明

6.3 手动切换PRIMARY节点

此时,还是选择mgr02作为PRIMARY节点,不会发生变化,除非手动执行切主操作:

[root@GreatSQL][(3306)]> select group_replication_set_as_primary('0fbb2cfd-c4d9-11eb-8747-525400e2078a');
+--------------------------------------------------------------------------+
| group_replication_set_as_primary('0fbb2cfd-c4d9-11eb-8747-525400e2078a') |
+--------------------------------------------------------------------------+
| Primary server switched to: 0fbb2cfd-c4d9-11eb-8747-525400e2078a         |
+--------------------------------------------------------------------------+

[root@GreatSQL][(3306)]> select * from performance_schema.replication_group_members;
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| CHANNEL_NAME              | MEMBER_ID                            |  MEMBER_HOST | MEMBER_PORT | MEMBER_STATE | MEMBER_ROLE | MEMBER_VERSION |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+
| group_replication_applier | 0fbb2cfd-c4d9-11eb-8747-525400e2078a | greatsql-mgr |        3306 | ONLINE      |  PRIMARY     | 8.0.25         |
| group_replication_applier | 1778e87a-c4d9-11eb-bf33-525400e2078a | greatsql-mgr |        3307 | ONLINE       | PRIMARY     | 8.0.25         |
| group_replication_applier | 1c8ebfa6-c4d9-11eb-80d1-525400e2078a | greatsql-mgr |        3308 | ONLINE       | SECONDARY   | 8.0.25         |
| group_replication_applier | bc037d6c-c4de-11eb-a5b8-525400e2078a | greatsql-mgr |        3309 | ONLINE       | SECONDARY   | 8.0.25         |
+---------------------------+--------------------------------------+--------------+-------------+--------------+-------------+----------------+

这就完成滚动升级以及再次切主的全部操作了。

至此,MGR集群的构建、添加新节点、滚动升级、切主等操作都已完成。

在单机多节点构建MGR集群,和在多机上的构建过程并无本质区别,大家可以自行操作一遍。

1
https://gitee.com/GreatSQL/GreatSQL-Doc.git
git@gitee.com:GreatSQL/GreatSQL-Doc.git
GreatSQL
GreatSQL-Doc
GreatSQL-Doc
master

Search