在Azure云上实现postgres主备切换

　　以下是工作上实现postgres主备切换功能所用到的代码和步骤，中间走了不少弯路，在此记录下。所用到的操作系统为centos 7.5，安装了两台服务器，hostname为VM7的为Master，VM8则为Slave。

　　1、安装pg10

　　vm7（Mater）,vm8（Slave）均需安装:

[root@springcloud-vm7 ~]# yum install –y https://download.postgresql.org/pub/repos/yum/reporpms/EL-7-x86_64/pgdg-redhat-repo-latest.noarch.rpm 
[root@springcloud-vm7 ~]# yum install postgresql10 -y
[root@springcloud-vm7 ~]# yum install postgresql10-server -y
[root@springcloud-vm7 ~]# systemctl enable postgresql-10
[root@springcloud-vm7 ~]# /usr/pgsql-10/bin/postgresql-10-setup initdb
[root@springcloud-vm7 ~]# systemctl start postgresql-10
[root@springcloud-vm7 ~]# systemctl status postgresql-10

　　其他环境设置

　　vm7（Mater）,vm8（Slave）均需安装:

# hosts
[root@springcloud-vm7 ~]# vi /etc/hosts
...
10.0.0.14 springcloud-vm7 vm7
10.0.0.15 springcloud-vm8 vm8

# disable selinux
[root@springcloud-vm7 ~]# sed -i '7s/=.*$/=disabled/' /etc/selinux/config
[root@springcloud-vm7 ~]# setenforce 0

# set timezone
[root@springcloud-vm7 ~]# timedatectl set-timezone Asia/Shanghai
# postgres⽤用户环境
[root@springcloud-vm7 ~]# su - postgres
-bash-4.2$ vi ~/.bash_profile
。。。
PATH=$PATH:/usr/pgsql-10/bin;export PATH
-bash-4.2$ . ~/.bash_profile

#注意：~/.bash_profile改后为：
PGDATA=/usr/local/pgsql/data
PATH=/usr/local/pgsql/bin:$PATH
export PGDATA PATH

　　安装nginx充当给azure做探测的服务（probe）：

[root@springcloud-vm7 ~]# yum install -y epel-release
[root@springcloud-vm7 ~]# yum install -y nginx
[root@springcloud-vm7 ~]# vi /etc/nginx/nginx.conf
...
listen 5999 default_server;   #修改监听端口在5999上

　　由于Azure上LB的floatingIP限制，它会将流量量导向probe成功的后端机器器，并且只能有⼀一台probe成功。

　　参考官⽅方例例⼦子，它是⽤用iptable来阻⽌止probe：https://github.com/Azure/azure-quickstart-templates/tree/master/haproxy-redundant-floatingip-ubuntu

　　azure的LSB设置：

　　注意：负载均衡的probe端⼝口设置成5999，floatingIP设置启⽤用

　　2、配置master

[root@springcloud-vm7 ~]# su – postgres

# 创建复制⽤用户
-bash-4.2$ psql -c "create role repl replication login password 'postgres'"

# 创建⼀一个slot
-bash-4.2$ psql -c "select pg_create_physical_replication_slot('slot_vm7')"

pg_create_physical_replication_slot
-------------------------------------
(slot_vm7,)
(1 row)

# 配置参数
-bash-4.2$ cd $PGDATA
-bash-4.2$ vi postgresql.conf
。。。
listen_addresses = '*'
archive_mode = on
archive_command = 'cp -n %p $PGDATA/arch/%f'
log_timezone = 'Asia/Shanghai'
timezone = 'Asia/Shanghai'
。。。
# 创建arch⽬目录
-bash-4.2$ mkdir $PGDATA/arch/

# 配置pg_hba.conf
-bash-4.2$ vi pg_hba.conf
。。。
host replication repl 10.0.0.0/24 md5

# 重启
-bash-4.2$ pg_ctl restart


注意：如果是首次安装，需要在防火墙中开放5432端口
#查看各端口网络连接情况
[root@springcloud-vm7 ~]# netstat –na

#安装iptables防火墙
[root@springcloud-vm7 ~]# yum install iptables-services 

#编辑iptables防火墙配置
[root@springcloud-vm7 ~]# vi /etc/sysconfig/iptables 
。。。
-A INPUT -m state --state NEW -m tcp -p tcp --dport 5432 -j ACCEPT

　　3、配置slave

 [root@springcloud-vm8 ~]# su – postgres

# 直接⽤用repl⽤用户备份到$PGDATA⽬目录
-bash-4.2$ rm -rf /var/lib/pgsql/10/data
-bash-4.2$ /usr/pgsql-10/bin/pg_basebackup -R -Pv -h vm7 -U repl -D $PGDATA
Password: #postgres

pg_basebackup: initiating base backup, waiting for checkpoint to complete
pg_basebackup: checkpoint completed
pg_basebackup: write-ahead log start point: 0/2000028 on timeline 1
pg_basebackup: starting background WAL receiver
24421/24421 kB (100%), 1/1 tablespace
pg_basebackup: write-ahead log end point: 0/20000F8
pg_basebackup: waiting for background process to finish streaming ...
pg_basebackup: base backup completed

# 修改recovery.conf⽂文件
-bash-4.2$ vi $PGDATA/recovery.conf

standby_mode = 'on'
primary_conninfo = 'user=repl password=postgres host=vm7 port=5432 sslmode=prefer sslcompression=1 krbsrvname=postgres target_session_attrs=any'
primary_slot_name = 'slot_vm7'
restore_command = 'cp $PGDATA/arch/%f %p'
archive_cleanup_command = 'pg_archivecleanup $PGDATA/arch %r'
recovery_target_timeline = 'latest'

# 重启
-bash-4.2$ pg_ctl restart

　　4、检查

#主库上检查下：
-bash-4.2$ psql -xc "select * from pg_stat_replication"
-[ RECORD 1 ]----+------------------------------
pid | 90512
usesysid | 16384
usename | repl
application_name | walreceiver
client_addr | 10.0.0.11
client_hostname |
client_port | 44464
backend_start | 2019-07-09 18:36:25.005202+08
backend_xmin |
state | streaming
sent_lsn | 0/F0004A0
write_lsn | 0/F0004A0
flush_lsn | 0/F0004A0
replay_lsn | 0/F0004A0
write_lag |
flush_lag |
replay_lag |
sync_priority | 0
sync_state | async
-bash-4.2$ pg_controldata
...
-bash-4.2$ psql -xc "select * from pg_replication_slots"
...

#被库上检查：
-bash-4.2$ psql -xc "select * from pg_stat_wal_receiver"
...
-bash-4.2$ /usr/pgsql-10/bin/pg_controldata
...
-bash-4.2$ psql -xc "select pg_is_in_recovery()"
-[ RECORD 1 ]-----+--
pg_is_in_recovery | t

　　5、安装keepalived

　　vm7（Mater）,vm8（Slave）均需安装:

[root@springcloud-vm7 data]# yum install keepalived -y
[root@springcloud-vm7 data]# systemctl enable keepalived
[root@springcloud-vm7 data]# systemctl start keepalived
[root@springcloud-vm7 data]# cd /etc/keepalived/
[root@springcloud-vm7 ~]# vi /etc/keepalived/keepalived.conf

! Configuration File for keepalived
global_defs {
notification_email {
admin@example.com
}
notification_email_from pg@example.com
smtp_server 127.0.0.1
smtp_connect_timeout 30
router_id PG_HA     #主备库需要一致
}

vrrp_script chk_pg_alived {
script "/sbin/ss -ntlp4 | grep :5432 > /dev/null" # 探测端⼝口判断数据库存活，1分钟失败则认为失败
interval 10
weight 20
fall 6
}

vrrp_instance VI_1 {
state MASTER # 主库上填MASTER, 备库上为BACKUP
interface eth0  # 填写当前网卡名称，可以用IP Ad命令查看
virtual_router_id 61
priority 100 #备库的优先级设为90
advert_int 3
!nopreempt
preempt_delay 60
unicast_src_ip 10.0.0.14 # 云主机只能使⽤用单播⽅方式，这⾥里里填本机ip
unicast_peer {
10.0.0.15 #另⼀台ip
}

authentication {
auth_type PASS
auth_pass 1111
}

virtual_ipaddress {
139.217.92.247  #虚拟ip
}

track_script {
chk_pg_alived
}

notify_master "/etc/keepalived/master.sh"
notify_backup "/etc/keepalived/backup.sh"
}

# 切换成master时会执⾏行行的脚本，通过判断数据库状态决定是否promote
[root@springcloud-vm7 keepalived]# vi master.sh

#!/bin/bash
export PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin

# allow probe from azure
systemctl restart nginx
dbstate=`su - postgres -c "psql -Atc 'select pg_is_in_recovery()'"`
if [ $dbstate != "t" ]; then
exit 0
fi

# promote the slave to master
su - postgres -c "/usr/pgsql-10/bin/pg_ctl promote"
sleep 5

echo "select pg_create_physical_replication_slot('slot_vm7')" | su – postgres -c "psql" #注意slot的名字主备库要相应修改⼀⼀对应（Slave中改为slot_vm8）


#保存后修改脚本文件权限
[root@springcloud-vm7 keepalived]# chmod 777 master.sh

#切换成slave时会执⾏行行的脚本
[root@springcloud-vm7 keepalived]# vi backup.sh

#!/bin/bash
export PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin
# block probe from azure
systemctl stop nginx
# check pg state
dbstate=`su - postgres -c "psql -Atc 'select pg_is_in_recovery()'"`
if [ $dbstate = "t" ]; then
exit 0
fi
# change master to slave
if [ ! -f /var/lib/pgsql/10/data/recovery.conf ] ; then
cat > /var/lib/pgsql/10/data/recovery.conf << EOF
standby_mode = 'on'
primary_conninfo = 'user=repl password=postgres host=vm8 port=5432 sslmode=prefer sslcompression=1 krbsrvname=postgres target_session_attrs=any'  #注意slot的名字主备库要相应修改⼀⼀对应（Slave中改为vm7）
primary_slot_name = 'slot_vm8' #注意slot的名字主备库要相应修改⼀⼀对应（Slave中改为slot_vm7）
restore_command = 'cp $PGDATA/arch/%f %p'
archive_cleanup_command = 'pg_archivecleanup $PGDATA/arch %r'
recovery_target_timeline = 'latest'
EOF
fi
sleep 60
su - postgres -c "/usr/pgsql-10/bin/pg_ctl stop"
systemctl start postgresql-10

#注意slot的名字主备库要相应修改⼀⼀对应

#保存后修改脚本文件权限
[root@springcloud-vm7 keepalived]# chmod 777 backup.sh