MGR新节点RECOVERING状态的分析与解决:caching_sha2_password验证插件的影响
* GreatSQL社区原创内容未经授权不得随意使用,转载请联系小编并注明来源。
起因
在GreatSQL社区上有一位用户提出了“手工构建MGR碰到的次节点一直处于recovering状态”,经过排查后,发现了是因为新密码验证插件caching_sha2_password
导致的从节点一直无法连接主节点,帖子地址:➥https://greatsql.cn/thread-420-2-1.html
复现
环境介绍
本文验证环境,以及本文所采用数据库为GreatSQL 8.0.32-24
$ cat /etc/system-release
Red Hat Enterprise Linux Server release 7.9 (Maipo)
$ uname -a
Linux gip 3.10.0-1160.el7.x86_64 #1 SMP Tue Aug 18 14:50:17 EDT 2020 x86_64 x86_64 x86_64 GNU/Linux
部署准备:
采用的是单机多实例的部署方式,如何部署单机多实例可以前往➥https://gitee.com/GreatSQL/GreatSQL-Manual/blob/master/6-oper-guide/6-6-multi-instances.md
IP | 端口 | 角色 |
---|---|---|
172.17.139.77 | 3306 | mgr01 |
172.17.139.77 | 3307 | mgr02 |
MGR有关配置参数:
#mgr settings
loose-plugin_load_add = 'mysql_clone.so'
loose-plugin_load_add = 'group_replication.so'
loose-group_replication_group_name = "aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaa1"
loose-group_replication_group_seeds = '172.17.139.77:33061,172.17.139.77:33071'
loose-group_replication_start_on_boot = ON
loose-group_replication_bootstrap_group = OFF
loose-group_replication_exit_state_action = READ_ONLY
loose-group_replication_flow_control_mode = "DISABLED"
loose-group_replication_single_primary_mode = ON
loose-group_replication_communication_max_message_size = 10M
loose-group_replication_transaction_size_limit = 3G
loose-group_replication_arbitrator = 0
loose-group_replication_single_primary_fast_mode = 0
loose-group_replication_request_time_threshold = 20000
report_host = "172.17.139.77"
MGR01节点配置如下:
[mysqld@mgr01]
datadir=/data/GreatSQL/mgr01
socket=/data/GreatSQL/mgr01/mysql.sock
port=3306
server_id=103306
log-error=/data/GreatSQL/mgr01/error.log
loose-group_replication_local_address= "172.17.139.77:33061"
MGR02节点配置如下:
[mysqld@mgr02]
datadir=/data/GreatSQL/mgr02
socket=/data/GreatSQL/mgr02/mysql.sock
port=3307
server_id=103317
log-error=/data/GreatSQL/mgr02/error.log
loose-group_replication_local_address= "172.17.139.77:33071"
启动MGR01实例、MGR02实例,并修改密码
#启动两个实例
$ systemctl restart greatsql@mgr01 &
$ systemctl restart greatsql@mgr02 &
# 获取初始化密码
$ grep root /data/GreatSQL/mgr01/error.log
$ grep root /data/GreatSQL/mgr02/error.log
# 登录数据库并修改密码
$ mysql -S /data/GreatSQL/mgr01/mysql.sock -uroot -p
greatsql> alter user root@'localhost' identified by 'GreatSQL@666';
$ mysql -S /data/GreatSQL/mgr02/mysql.sock -uroot -p
greatsql> alter user root@'localhost' identified by 'GreatSQL@666';
检查两个实例是否正确加载group_replicaiton
插件
greatsql> show plugins;
+----------------------------------+----------+--------------------+----------------------+---------+
| Name | Status | Type | Library | License |
+----------------------------------+----------+--------------------+----------------------+---------+
| group_replication | ACTIVE | GROUP REPLICATION | group_replication.so | GPL |
+----------------------------------+----------+--------------------+----------------------+---------+
没有加载的话可以手动加载这个plugin
greatsql> install plugin group_replication soname 'group_replication.so';
搭建MGR
接下来就可以手工搭建MGR,流程如下可参考安装部署MGR集群 | 深入浅出MGR
➥https://gitee.com/GreatSQL/GreatSQL-Doc/blob/master/deep-dive-mgr/deep-dive-mgr-03.md
MGR01实例操作:
greatsql> set session sql_log_bin=0;
# 特别注意下面因为8.0.4版本开始使用的默认是“caching_sha2_password”,所以这样创建会采用最新的身份认证插件
greatsql> create user repl@'%' identified by 'GreatSQL@666';
greatsql> GRANT BACKUP_ADMIN, REPLICATION SLAVE ON *.* TO `repl`@`%`;
greatsql> set session sql_log_bin=1;
greatsql> CHANGE MASTER TO MASTER_USER='repl', MASTER_PASSWORD='GreatSQL@666' FOR CHANNEL 'group_replication_recovery';
接下来即可启动MGR集群:
greatsql> set global group_replication_bootstrap_group=ON;
greatsql> start group_replication;
greatsql> select * from performance_schema.replication_group_members\G
*************************** 1. row ***************************
CHANNEL_NAME: group_replication_applier
MEMBER_ID: 2920447e-35bf-11ee-89a5-00163e566da1
MEMBER_HOST: 172.17.139.77
MEMBER_PORT: 3306
MEMBER_STATE: ONLINE
MEMBER_ROLE: PRIMARY
MEMBER_VERSION: 8.0.32
MEMBER_COMMUNICATION_STACK: XCom
MGR02实例操作:
greatsql> set session sql_log_bin=0;
greatsql> create user repl@'%' identified by 'GreatSQL@666';
greatsql> GRANT BACKUP_ADMIN, REPLICATION SLAVE ON *.* TO `repl`@`%`;
greatsql> set session sql_log_bin=1;
greatsql> CHANGE MASTER TO MASTER_USER='repl', MASTER_PASSWORD='GreatSQL@666' FOR CHANNEL 'group_replication_recovery';
greatsql> start group_replication;
Query OK, 0 rows affected (5.39 sec)
此时创建的用户采用的都是caching_sha2_password
身份认证插件
greatsql> SELECT USER,PLUGIN FROM mysql.`user` ;
+------------------+-----------------------+
| USER | PLUGIN |
+------------------+-----------------------+
| repl | caching_sha2_password |
| mysql.infoschema | caching_sha2_password |
| mysql.session | caching_sha2_password |
| mysql.sys | caching_sha2_password |
| root | caching_sha2_password |
+------------------+-----------------------+
虽然启动MGR成功,但是查看下节点状态:
greatsql> select * from performance_schema.replication_group_members\G
*************************** 1. row ***************************
CHANNEL_NAME: group_replication_applier
MEMBER_ID: 2920447e-35bf-11ee-89a5-00163e566da1
MEMBER_HOST: 172.17.139.77
MEMBER_PORT: 3306
MEMBER_STATE: ONLINE
MEMBER_ROLE: PRIMARY
MEMBER_VERSION: 8.0.32
MEMBER_COMMUNICATION_STACK: XCom
*************************** 2. row ***************************
CHANNEL_NAME: group_replication_applier
MEMBER_ID: 2a4f068b-35bf-11ee-9504-00163e566da1
MEMBER_HOST: 172.17.139.77
MEMBER_PORT: 3307
MEMBER_STATE: RECOVERING
MEMBER_ROLE: SECONDARY
MEMBER_VERSION: 8.0.32
MEMBER_COMMUNICATION_STACK: XCom
2 rows in set (0.00 sec)
此时节点一直处于RECOVERING
状态,查看mgr02实例的错误日志如下:
2023-08-08T08:00:47.034870Z 42 [ERROR] [MY-010584] [Repl] Slave I/O for channel 'group_replication_recovery': error connecting to master 'repl@172.17.139.77:3306' - retry-time: 60 retries: 1 message: Authentication plugin 'caching_sha2_password' reported error: Authentication requires secure connection. Error_code: MY-002061
2023-08-08T08:00:47.037631Z 35 [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.'
2023-08-08T08:00:47.037671Z 35 [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.'
这是由于caching_sha2_password
是 MySQL 8.0.4 引入的一个新的身份验证插件,caching_sha2_password
对密码安全性要求更高,要求用户认证过程中在网络传输的密码是加密的,所以导致的这个问题的出现,caching_sha2_password
的介绍可以看社区文章“浅谈 MySQL 新的身份验证插件 caching_sha2_password”
解决方式
1、采用旧密码验证插件
旧的身份验证插件mysql_native_password
,mysql_native_password
的特点是不需要加密的连接。该插件验证速度特别快,但是不够安全,只需要更改创建用户的语句
create user repl@'%' identified with mysql_native_password by 'GreatSQL@666';
旧密码验证插件容易被破解,如果有 GreatSQL 服务要公网上使用,建议还是尽量使用 caching_sha2_password
作为认证插件
2、启用group_replication_recovery_get_public_key
设置 group_replication_recovery_get_public_key=ON
可以确保从节点在连接到主节点时能够获取所需的公钥,从而允许安全连接并成功进行身份验证,避免了连接错误和身份验证问题。
手册中也有明确说明:
18.6.3.1.1 Replication User With The Caching SHA-2 Authentication Plugin
By default, users created in MySQL 8 use Section 6.4.1.2, “Caching SHA-2 Pluggable Authentication”. If the replication user you configure for distributed recovery uses the caching SHA-2 authentication plugin, and you are not using SSL for distributed recovery connections, RSA key-pairs are used for password exchange. For more information on RSA key-pairs, see Section 6.3.3, “Creating SSL and RSA Certificates and Keys”.
In this situation, you can either copy the public key of the to the joining member, or configure the donors to provide the public key when requested. The more secure approach is to copy the public key of the replication user account to the joining member. Then you need to configure the group_replication_recovery_public_key_path system variable on the joining member with the path to the public key for the replication user account. rpl_user
The less secure approach is to set group_replication_recovery_get_public_key=ON on donors so that they provide the public key of the replication user account to joining members. There is no way to verify the identity of a server, therefore only set group_replication_recovery_get_public_key=ON when you are sure there is no risk of server identity being compromised, for example by a man-in-the-middle attack.
可以看到,当确认环境安全以及没人任何人攻击集群时,如果不配置ssl,可以最低配置group_replication_recovery_get_public_key=ON
来在请求复制用户密钥时给公钥
3、为组复制通道启用SSL支持
以下操作方法仅使用于 GreatSQL/MySQL 8.0.27 版本及以上
更安全的方法是将repl用户所需的公钥文件复制到joiner节点的Server所在主机中。然后,在joiner节点的Server中配置group_replication_recovery_public_key_path
系统变量,指定rpl_user用户所需的公钥文件路径。
使用caching_sha2_password
插件身份验证会在数据目录下生成如下两个RSA文件:
private_key.pem
public_key.pem
private_key.pem
:RSA私钥public_key.pem
:RSA公钥
对于 MGR ,如果设置 group_replication_ssl_mode=DISABLED
必须使用下面的变量来指定 RSA 公钥,否则报错:
group_replication_recovery_get_public_key
:向服务端请求 RSA 公钥;group_replication_recovery_public_key_path
:指定本地 RSA 公钥文件;
指定本地RSA公钥,首先需要全局MGR配置开启SSL
[mysqld]
# 开启use_ssl,指定组成员之间的组复制分布式恢复连接是否应使用 SSL
loose-group_replication_recovery_use_ssl=ON
进入MGR01实例配置
greatsql> set session sql_log_bin=0;
# 此时就可以使用“caching_sha2_password”身份认证插件
greatsql> create user repl@'%' identified by 'GreatSQL@666';
greatsql> GRANT BACKUP_ADMIN, REPLICATION SLAVE ON *.* TO `repl`@`%`;
greatsql> set session sql_log_bin=1;
greatsql> CHANGE MASTER TO MASTER_USER='repl', MASTER_PASSWORD='GreatSQL@666' FOR CHANNEL 'group_replication_recovery';
启动MGR01实例的MGR集群
greatsql> set global group_replication_bootstrap_group=ON;
greatsql> start group_replication;
greatsql> select * from performance_schema.replication_group_members\G
*************************** 1. row ***************************
CHANNEL_NAME: group_replication_applier
MEMBER_ID: 35b653d2-3658-11ee-93c9-00163e566da1
MEMBER_HOST: 172.17.139.77
MEMBER_PORT: 3306
MEMBER_STATE: ONLINE
MEMBER_ROLE: PRIMARY
MEMBER_VERSION: 8.0.32
MEMBER_COMMUNICATION_STACK: XCom
启动成功后,需要把MGR01节点的RSA公钥
拷贝到MGR02节点上,因为MGR02也会生成此公钥,所以最好创建一个文件夹
$ mkdir mgr01_key
$ chown mysql:mysql mgr01_key/
# 将public_key.pem移动到MGR02
$ mv /data/GreatSQL/mgr01/public_key.pem /data/GreatSQL/mgr02/mgr01_key/
当然,如果有多个节点,也需要把主节点的RSA公钥移动到各个节点上
MGR02节点操作
greatsql> set session sql_log_bin=0;
greatsql> create user repl@'%' identified by 'GreatSQL@666';
greatsql> GRANT BACKUP_ADMIN, REPLICATION SLAVE ON *.* TO `repl`@`%`;
greatsql> set session sql_log_bin=1;
greatsql> CHANGE MASTER TO MASTER_USER='repl', MASTER_PASSWORD='GreatSQL@666' FOR CHANNEL 'group_replication_recovery';
# 此命令设置完成后,最好写进my.cnf文件中持久化
greatsql> set global group_replication_recovery_public_key_path = "/data/GreatSQL/mgr02/mgr01key/public_key.pem";
greatsql> start group_replication;
greatsql> select * from performance_schema.replication_group_members\G
*************************** 1. row ***************************
CHANNEL_NAME: group_replication_applier
MEMBER_ID: 35b653d2-3658-11ee-93c9-00163e566da1
MEMBER_HOST: 172.17.139.77
MEMBER_PORT: 3306
MEMBER_STATE: ONLINE
MEMBER_ROLE: PRIMARY
MEMBER_VERSION: 8.0.32
MEMBER_COMMUNICATION_STACK: XCom
*************************** 2. row ***************************
CHANNEL_NAME: group_replication_applier
MEMBER_ID: aa031fb9-365a-11ee-9925-00163e566da1
MEMBER_HOST: 172.17.139.77
MEMBER_PORT: 3307
MEMBER_STATE: ONLINE
MEMBER_ROLE: SECONDARY
MEMBER_VERSION: 8.0.32
MEMBER_COMMUNICATION_STACK: XCom
可以看到双节点ONLINE
,新加入的节点不会一直是RECOVERING
状态
总结
新身份验证插件caching_sha2_password
安全度相比其他的身份验证插件,既解决安全性问题又解决性能问题,建议使用新密码验证插件。
也感谢社区用户指出GreatSQL社区文档中的不足,并给予用户金币奖励,同时欢迎大家来GreatSQL社区捉虫~
《深入浅出MGR》视频课程
戳此小程序即可直达B站
<文 章 推 荐>
GreatSQL茶话会07 | 你常用的好用的数据库周边工具有哪些?
9.16 北京 | GreatSQL 邀你参加 3306Π 数据库朋友圈线下 meetup
<关于 GreatSQL>
GreatSQL 是适用于金融级应用的国内自主开源数据库,具备高性能、高可靠、高易用性、高安全等多个核心特性,可以作为 MySQL 或 Percona Server 的可选替换,用于线上生产环境,且完全免费并兼容 MySQL 或 Percona Server。
https://greatsql.cn/thread-54-1-1.html
https://greatsql.cn/thread-100-1-1.html
(对文章有疑问或者有独到见解都可以去社区官网提出或分享哦~)
微信&QQ群:
可扫码添加GreatSQL社区助手微信好友,发送验证信息“加群”加入GreatSQL/MGR交流微信群,亦可直接扫码加入GreatSQL/MGR交流QQ群。
微信
想看更多技术好文,点个“在看”吧!