1、准備好3台CentOS系統機器
1）修改hostname為mysql1、mysql2、mysql3，修改/etc/hostname文件，將里面內容改掉即可
vi /etc/hostname
2）修改/etc/hosts文件，配置映射
vi /etc/hosts
配置如下：
172.16.43.141 mysql1 172.16.43.142 mysql2 172.16.43.143 mysql3
3）禁用防火牆或開放3306和33061端口
systemctl stop firewalld.service
systemctl disable firewalld.service
4）禁用SELINUX，修改/etc，SELINUX=enforcing改為SELINUX=disabled（為啥這樣做不是很清楚，不操作的話后面無法創建集群）
vi /etc/selinux/config
5）重啟服務器，讓配置生效。
2、安裝MySQL
詳細安裝過程見：https://www.cnblogs.com/zhi-leaf/p/5993676.html
mysql80-community-release-el7-3.noarch.rpm默認安裝的MySQL8，如果想安裝MySQL5.7，需要修改/etc/yum.repos.d/mysql-community.repo，將mysql57和mysql80的enabled值對調一下
vi /etc/yum.repos.d/mysql-community.repo
3、安裝mysql-shell和mysql-router
yum install mysql-community-server -y yum install mysql-shell -y yum install mysql-router -y
4、查看mysqld配置是否滿足集群
mysqlsh --log-level=DEBUG3 dba.checkInstanceConfiguration('root@mysql1:3306') # 檢測mysql服務是否滿足集群條件
5、根據上面的檢查，我們修改對應的配置（所有MySQL服務節點都需要修改）
1）退出MySQL JS，重新登陸到具體服務
mysqlsh --uri root@localhost:3306
2）執行dba.configureLocalInstance()，自動修改配置（配置修改完成后一定要重啟mysqld服務）
dba.configureLocalInstance('root@localhost:3306')
dba.checkInstanceConfiguration()
再次使用dba.checkInstanceConfiguration()，當時檢查結果status為ok時，說明配置已修改完成，MySQL服務符合集群條件，這時只需重啟mysqld服務就可以了
service mysqld restart
說明一點：mysql8.0直接修改運行參數，而ysql5.7會將修改的配置寫入/etc/my.cnf文件，他在my.cnf中增加的配置信息如下圖（這些修改項正是 dba.checkInstanceConfiguration() 檢查的修改項，我們直接復制過去也是可以的）：
binlog_checksum = NONE enforce_gtid_consistency = ON gtid_mode = ON log_bin log_slave_updates = ON master_info_repository = TABLE relay_log_info_repository = TABLE server_id = 3293780721 transaction_write_set_extraction = XXHASH64
6、mysql服務重啟完成后，我們就可以創建集群，並添加2個節點（只需在一個節點操作）
mysqlsh --uri root@localhost:3306 # 進入mysqlsh客戶端
var cluster = dba.createCluster('dev') cluster.addInstance('root@mysql2:3306'); cluster.addInstance('root@mysql3:3306');
再注意一點：mysql57在創建集群和添加節點的時候會出現警告：
WARNING: Instance 'mysql1:3306' cannot persist configuration since MySQL version 5.7.29 does not support the SET PERSIST command (MySQL version >= 8.0.11 required). Please use the <Dba>.configureLocalInstance() command locally to persist the changes.
這個警告沒關系，他告訴我們做集群修改的配置沒有持久化，我們再次在每個節點都執行dba.configureLocalInstance()就可以了，他會一大堆group_replication配置信息寫入到/etc/my.cnf。
新添加的配置項如下（我對比了一下，只有group_replication_group_seeds和group_replication_local_address2個參數不一致）：
auto_increment_increment = 1 auto_increment_offset = 2 group_replication_allow_local_disjoint_gtids_join = OFF group_replication_allow_local_lower_version_join = OFF group_replication_auto_increment_increment = 7 group_replication_bootstrap_group = OFF group_replication_components_stop_timeout = 31536000 group_replication_compression_threshold = 1000000 group_replication_enforce_update_everywhere_checks = OFF group_replication_exit_state_action = READ_ONLY group_replication_flow_control_applier_threshold = 25000 group_replication_flow_control_certifier_threshold = 25000 group_replication_flow_control_mode = QUOTA group_replication_force_members = group_replication_group_name = eb782af9-732a-11ea-9cee-000c290200c6 group_replication_group_seeds = mysql1:33061,mysql3:33061 group_replication_gtid_assignment_block_size = 1000000 group_replication_ip_whitelist = AUTOMATIC group_replication_local_address = mysql2:33061 group_replication_member_weight = 50 group_replication_poll_spin_loops = 0 group_replication_recovery_complete_at = TRANSACTIONS_APPLIED group_replication_recovery_reconnect_interval = 60 group_replication_recovery_retry_count = 10 group_replication_recovery_ssl_ca = group_replication_recovery_ssl_capath = group_replication_recovery_ssl_cert = group_replication_recovery_ssl_cipher = group_replication_recovery_ssl_crl = group_replication_recovery_ssl_crlpath = group_replication_recovery_ssl_key = group_replication_recovery_ssl_verify_server_cert = OFF group_replication_recovery_use_ssl = ON group_replication_single_primary_mode = ON group_replication_ssl_mode = REQUIRED group_replication_start_on_boot = ON group_replication_transaction_size_limit = 0 group_replication_unreachable_majority_timeout = 0 super_read_only = ON
7、集群安裝完成，我們可以查看集群的安裝狀態
var cluster = dba.getCluster()
cluster.status()
可以看出，mysql1節點為讀寫模式，mysql2和mysql3節點為讀模式（只能查詢，不能執行建表、新增數據等操作）
8、集群驗證
關閉mysql1節點，幾秒鍾后，集群primary節點從mysql1切換到了msql3，並且mysql3變成了讀寫模式
更多安裝說明：https://dev.mysql.com/doc/refman/8.0/en/mysql-innodb-cluster-creating.html
遇到的問題：
問題1：Dba.createCluster: dba.createCluster: Unable to create cluster. The instance 'mysql1:3306' has a populated Metadata schema and belongs to that Metadata. Use either dba.dropMetadataSchema() to drop the schema, or dba.rebootClusterFromCompleteOutage() to reboot the cluster from complete outage. (RuntimeError)
解決：執行dba.rebootClusterFromCompleteOutage()
問題2：Cluster.addInstance: Cannot add an instance with the same server UUID (4245c0b6-7270-11ea-929d-000c292b8420) of an active member of the cluster 'mysql1:3306'. Please change the server UUID of the instance to add, all members must have a unique server UUID. (RuntimeError)
解決：我是虛擬機復制的，所有MySQL服務的server_uuid和server_id是一樣的，我們必須修改，servier_uuid需要修改/var/lib/mysql/auto.cnf文件，server_id通過sql語句設置。資料參考：https://dev.mysql.com/doc/refman/8.0/en/replication-options.html
vi /var/lib/mysql/auto.cnf // 修改server_uuid
set global server_id = 549754138; // 修改server_id
問題3：Unable to start Group Replication for instance 'mysql3:3306'. Please check the MySQL server error log for more information.
Cluster.addInstance: Group Replication failed to start: MySQL Error 3096 (HY000): mysql3:3306: The START GROUP_REPLICATION command failed as there was an error when initializing the group communication layer. (RuntimeError)
解決：mysql3主機的hosts配置錯了。
問題4：Dba.getCluster: This function is not available through a session to a standalone instance (metadata exists, instance belongs to that metadata, but GR is not active) (RuntimeError)
解決：解決：執行dba.rebootClusterFromCompleteOutage()