Greenplum——Master自动fail-over高可用方案

Greenplum——Master自动fail-over高可用方案

在 linux2.4 内核以后，已经完全内置了 LVS 的各个功能模块。
当内核版本小于2.5，需要把keepalived注册为系统服务

##uname -a查看内核版本信息
uname -a

安装
yum方式

安装依赖包
[root@localhost ~]# yum install -y curl gcc openssl-devel libnl3-devel net-snmp-devel
[root@localhost ~]# yum install -y keepalived

可能报错：ImportError: No module named site
1.报错原因
Centos自带的python版本为2.7, 升级后yum与高版本的python不兼容

2.解决方法
在终端输入:
unset PYTHONPATH
unset PYTHONHOME

keepalived+greenplum+虚拟ip方案

mastet配置

! Configuration File for keepalived

global_defs {
   router_id mdw
}
vrrp_script chk_gp {
script"/home/gpadmin/bin/check_gp.sh"
interval 5
}
vrrp_instance VI_1 {
    state MASTER
    nopreempt
    interface eno16777736
    virtual_router_id 51
    priority 110
    advert_int 1
    authentication {
        auth_type PASS
        auth_pass 123456
    }
    virtual_ipaddress {
        172.20.10.100
    }
track_script
 
{
chk_gp
}
notify /home/gpadmin/bin/keep.sh
}

standby

! Configuration File for keepalived

global_defs {
   router_id smdw
}
vrrp_script chk_gp {
script"/home/gpadmin/bin/check_gp.sh"
interval 5
}
vrrp_instance VI_1 {
    state BACKUP
    interface eno16777736
    virtual_router_id 51
    priority 105
    advert_int 1
    authentication {
        auth_type PASS
        auth_pass 123456
    }
    virtual_ipaddress {
        172.20.10.100
    }
track_script
 
{
chk_gp
}
notify /home/gpadmin/bin/keep.sh
}

其他相关脚本

1. check_gp.sh

#!/bin/sh
CHECK_PROCESS=`ps aux | grep postgres | grep 5432 | wc -l`
if [ $CHECK_PROCESS -eq 0 ]
then
echo “GP is stopped”
sleep 2
CHECK_PROCESS=`ps aux | grep postgres | grep 5432 |wc -l`
if [ $CHECK_PROCESS -eq 0 ]
then
	exit 1
else
	exit 0
#systemctl stop keepalived.service
fi
fi

2. keep.sh

#!/bin/bash
TYPE=$1
NAME=$2
STATE=$3
case $STATE in
MASTER) /bin/bash /home/gpadmin/bin/runner.sh >>/home/gpadmin/bin/keep.data;;
BACKUP) echo "Executing as BACKUP date" >> /home/gpadmin/bin/keep.data;;
FAULT) echo "Executing as FAULT date" >> /home/gpadmin/bin/keep.data exit 0;;
*) echo "Unknown state <<<<<<<<<<<<<<<<<<< date" >> /home/gpadmin/bin/keep.data exit 1;;
esac

3. runner.sh

#!/bin/sh
username=`whoami`
if [ "$username" = "gpadmin" ]; then
  # /etc/keepalived/bin/monitor.sh
    /home/gpadmin/bin/activateGPstandby.sh
else
  # /bin/su - gpadmin -c /etc/keepalived/bin/monitor.sh
  /bin/su - gpadmin -c /home/gpadmin/bin/activateGPstandby.sh
fi

4.activateGPstandby.sh

source /usr/local/greenplum-db-6.19.3/greenplum_path.sh
export export MASTER_DATA_DIRECTORY=/home/gpadmin/data/master/gpseg-1
export PGPORT=5432
echo "Activating MASTER on smdw date" >> /home/gpadmin/status
gpactivatestandby -a -d $MASTER_DATA_DIRECTORY

以上脚本可实现MASTER挂了之后服务访问BACKUP。此时BACKUP作为MASTER使用，当原MASTER具备能力时，将其挂载到原BACKUP节点上，作为原BACKUP的BACKUP。当原BACKUP挂掉之后，又可以访问原MASTER。

nopreempt和weight可以重点关注。动态权重weight在此场景下并不适用。动态权重可以在keepalive+nginx高可用和负载均衡场景下使用，具体可查看其他文章。