当前位置：首页 > news >正文

PolarDB 高可用集群搭建

news 2026/4/18 19:58:18

PolarDB 高可用集群搭建

准备三台机器（注意每个节点需配置一致）：
192.168.142.23
192.168.142.24
192.168.142.25

1、关闭透明大页
cat /sys/kernel/mm/transparent_hugepage/enabled
如返回结果为always madvise [never]则表示已关闭透明大页。
否则执行以下操作以关闭透明大页。
echo never > /sys/kernel/mm/transparent_hugepage/enabled
关闭防火墙：
systemctl disable firewalld
systemctl stop firewalld

2、安装主程序
[root@standby PolarDB]# yum -y install t-polardbx-engine-8.4.19-20250825_17558853.el7.x86_64.rpm
Loaded plugins: ulninfo
Examining t-polardbx-engine-8.4.19-20250825_17558853.el7.x86_64.rpm: t-polardbx-engine-8.4.19-20250825_17558853.el7.x86_64
Marking t-polardbx-engine-8.4.19-20250825_17558853.el7.x86_64.rpm to be installed
Resolving Dependencies
--> Running transaction check
---> Package t-polardbx-engine.x86_64 0:8.4.19-20250825_17558853.el7 will be installed
--> Finished Dependency Resolution

Dependencies Resolved

=====================================================================================================================================================================================
Package Arch Version Repository Size
=====================================================================================================================================================================================
Installing:
t-polardbx-engine x86_64 8.4.19-20250825_17558853.el7 /t-polardbx-engine-8.4.19-20250825_17558853.el7.x86_64 2.8 G

Transaction Summary
=====================================================================================================================================================================================
Install 1 Package

Total size: 2.8 G
Installed size: 2.8 G
Downloading packages:
Running transaction check
Running transaction test
Transaction test succeeded
Running transaction
Installing : t-polardbx-engine-8.4.19-20250825_17558853.el7.x86_64 1/1
Copying /u01/xcluster80_20250825_current to /u01/xcluster80
Copying /u01/xcluster80_20250825_current to /u01/xcluster80_20250825
Verifying : t-polardbx-engine-8.4.19-20250825_17558853.el7.x86_64 1/1

Installed:
t-polardbx-engine.x86_64 0:8.4.19-20250825_17558853.el7

Complete!
注：安装完成后，默认会创建/u01/目录，可以在u01目录下看到对应的数据库二进制文件，如下。
[root@standby PolarDB]# cd /u01/
[root@standby u01]# ll
drwxr-xr-x 11 root root 4096 Apr 14 11:59 xcluster80
drwxr-xr-x 11 root root 4096 Apr 14 11:59 xcluster80_20250825

3、创建并切换到 polardbx 用户
useradd -ms /bin/bash polardbx
echo "polardbx:polardbx" | chpasswd
echo "polardbx ALL=(ALL) NOPASSWD: ALL" >> /etc/sudoers
su - polardbx

4、配置初始化参数
确认当前操作系统用户是 polardbx 用户，而不是 root 用户。

准备文件目录（建议将数据存放到单独的存储路径下）：
mkdir -p /u01/polardbx-engine
chown -R polardbx:polardbx /u01/polardbx-engine
su - polardbx
cd /u01/polardbx-engine && mkdir log mysql run data tmp

准备配置文件（/u01/polardbx-engine/my.cnf），创建文件 touch my.cnf，并粘贴如下文本：
节点一：
$ vi my.cnf
[mysqld]
basedir = /u01/xcluster80
datadir = /u01/polardbx-engine/data
tmpdir = /u01/polardbx-engine/tmp
socket = /u01/polardbx-engine/tmp/mysql.sock
log_error = /u01/polardbx-engine/log/alert.log

port = 4886
default_authentication_plugin = mysql_native_password
lower_case_table_names = 1
gtid_mode = ON
enforce_gtid_consistency = ON
log_bin = 1
sync_binlog=1
innodb_flush_log_at_trx_commit=1
log_error_verbosity = 3
mysqlx=0
enable_polarx_rpc=0
loose-mysqlx-ssl=DISABLED
ssl=0

cluster_id = 1749112302
cluster_info = 192.168.142.23:14886@1
server_id = 1001

[mysqld_safe]
pid_file = /u01/polardbx-engine/run/mysql.pid

节点二：
$ vi my.cnf
[mysqld]
basedir = /u01/xcluster80
datadir = /u01/polardbx-engine/data
tmpdir = /u01/polardbx-engine/tmp
socket = /u01/polardbx-engine/tmp/mysql.sock
log_error = /u01/polardbx-engine/log/alert.log

cluster_id = 1749112302
cluster_info = 192.168.142.24:14886@2
server_id = 1002

[mysqld_safe]
pid_file = /u01/polardbx-engine/run/mysql.pid

节点三：
$ vi my.cnf
[mysqld]
basedir = /u01/xcluster80
datadir = /u01/polardbx-engine/data
tmpdir = /u01/polardbx-engine/tmp
socket = /u01/polardbx-engine/tmp/mysql.sock
log_error = /u01/polardbx-engine/log/alert.log

cluster_id = 1749112302
cluster_info = 192.168.142.25:14886@3
server_id = 1003

[mysqld_safe]
pid_file = /u01/polardbx-engine/run/mysql.pid

5、高可用集群初始化
每个节点依次按如下步骤初始化、启动：
节点一：
/u01/xcluster80/bin/mysqld --defaults-file=my.cnf --cluster-info='192.168.142.23:14886;192.168.142.24:14886;192.168.142.25:14886@1' --initialize-insecure
/u01/xcluster80/bin/mysqld_safe --defaults-file=my.cnf &

节点二：
/u01/xcluster80/bin/mysqld --defaults-file=my.cnf --cluster-info='192.168.142.23:14886;192.168.142.24:14886;192.168.142.25:14886@2' --initialize-insecure
/u01/xcluster80/bin/mysqld_safe --defaults-file=my.cnf &

节点三：
/u01/xcluster80/bin/mysqld --defaults-file=my.cnf --cluster-info='192.168.142.23:14886;192.168.142.24:14886;192.168.142.25:14886@3' --initialize-insecure
/u01/xcluster80/bin/mysqld_safe --defaults-file=my.cnf &

参数说明：
--cluster-info，其中的格式为 [host1]:[port1];[host2]:[port2];[host3]:[port3]@[idx] ，不同的机器只有 [idx] 不同，[idx] 也反映了该机器是第几个 [host][port]，请根据实际机器的 ip 修改该配置项。
比如：192.168.142.23/192.168.142.24/192.168.142.25 组成了Paxos多副本，在192.168.142.24上配置--cluster-info='192.168.142.23:14886;192.168.142.24:14886;192.168.142.25:14886@2'，其中@2代表192.168.142.24是在paxos集群配置里使用第二个节点的ip和端口，注意参数的差异。

6、登录集群数据库验证状态
/u01/xcluster80/bin/mysql -h127.0.0.1 -P4886 -uroot
查询本机的paxos角色
mysql> SELECT * FROM INFORMATION_SCHEMA.ALISQL_CLUSTER_LOCAL\G

查询集群所有机器的paxos角色（只有在Leader节点查询才会返回数据）
mysql> SELECT * FROM INFORMATION_SCHEMA.ALISQL_CLUSTER_GLOBAL \G
*************************** 1. row ***************************
SERVER_ID: 1
IP_PORT: 192.168.142.23:14886
MATCH_INDEX: 1
NEXT_INDEX: 0
ROLE:Leader
HAS_VOTED: Yes
FORCE_SYNC: No
ELECTION_WEIGHT: 5
LEARNER_SOURCE: 0
APPLIED_INDEX: 1
PIPELINING: No
SEND_APPLIED: No
INSTANCE_TYPE: Normal
DISABLE_ELECTION: No
SERVER_IP:
SERVER_PORT: 0
*************************** 2. row ***************************
SERVER_ID: 2
IP_PORT: 192.168.142.24:14886
MATCH_INDEX: 1
NEXT_INDEX: 2
ROLE: Follower
HAS_VOTED: Yes
FORCE_SYNC: No
ELECTION_WEIGHT: 5
LEARNER_SOURCE: 0
APPLIED_INDEX: 1
PIPELINING: Yes
SEND_APPLIED: No
INSTANCE_TYPE: Normal
DISABLE_ELECTION: No
SERVER_IP:
SERVER_PORT: 0
*************************** 3. row ***************************
SERVER_ID: 3
IP_PORT: 192.168.142.25:14886
MATCH_INDEX: 1
NEXT_INDEX: 2
ROLE: Follower
HAS_VOTED: No
FORCE_SYNC: No
ELECTION_WEIGHT: 5
LEARNER_SOURCE: 0
APPLIED_INDEX: 1
PIPELINING: Yes
SEND_APPLIED: No
INSTANCE_TYPE: Normal
DISABLE_ELECTION: No
SERVER_IP:
SERVER_PORT: 0
3 rows in set (0.00 sec)

注意：Paxos三副本在逐台启动时，刚启动第一台时，会因为不满足Paxos多数派，无法产生选主结果，此时数据库无法登录。至少需要第二个节点加入，并成功选主，才能登录数据库。

创建数据库、表，验证数据：
mysql> create database zzh;
Query OK, 1 row affected (0.01 sec)

mysql> use zzh;
Database changed
mysql> create table test (id int);
Query OK, 0 rows affected (0.02 sec)

mysql> insert into test values(1),(2);
Query OK, 2 rows affected (0.11 sec)
Records: 2 Duplicates: 0 Warnings: 0

在 Leader 上查询集群的状态：
mysql> SELECT SERVER_ID,IP_PORT,MATCH_INDEX,ROLE,APPLIED_INDEX FROM INFORMATION_SCHEMA.ALISQL_CLUSTER_GLOBAL ;
+-----------+----------------------+-------------+----------+---------------+
| SERVER_ID | IP_PORT | MATCH_INDEX | ROLE | APPLIED_INDEX |
+-----------+----------------------+-------------+----------+---------------+
| 1 | 192.168.142.23:14886 | 4 | Leader | 4 |
| 2 | 192.168.142.24:14886 | 4 | Follower | 4 |
| 3 | 192.168.142.25:14886 | 4 | Follower | 4 |
+-----------+----------------------+-------------+----------+---------------+
3 rows in set (0.00 sec)
注：其中 APPLIED_INDEX 都是 4 ，说明数据目前Paxos三节点上的Log Index是完全一致的。

7、验证重新选主
kill Leader 主节点进程，注意观察日志：
kill -9 $(pgrep -x mysqld)

kill 后集群会选出新的 Leader，并且 mysqld_safe 会立马重新拉起 mysqld 进程如下：
... ...
2026-04-17T02:38:58.502718Z mysqld_safe Rename corefile from to
2026-04-17T02:38:58.536787Z mysqld_safe Number of processes running now: 0
2026-04-17T02:38:58.547884Z mysqld_safe mysqld restarted

再次查看集群，Leader 变成了 192.168.142.24 节点，23节点变成了 Follower 如下：
mysql> SELECT SERVER_ID,IP_PORT,MATCH_INDEX,ROLE,APPLIED_INDEX FROM INFORMATION_SCHEMA.ALISQL_CLUSTER_GLOBAL ;
+-----------+----------------------+-------------+----------+---------------+
| SERVER_ID | IP_PORT | MATCH_INDEX | ROLE | APPLIED_INDEX |
+-----------+----------------------+-------------+----------+---------------+
| 1 | 192.168.142.23:14886 | 5 | Follower | 5 |
| 2 | 192.168.142.24:14886 | 5 | Leader | 5 |
| 3 | 192.168.142.25:14886 | 5 | Follower | 5 |
+-----------+----------------------+-------------+----------+---------------+
3 rows in set (0.00 sec)

至此：polardb 高可用集群搭建并验证完成。

查看全文

http://www.jsqmd.com/news/662399/