HA高可用集群
实验环境
node1:192.168.0.1
node2:192.168.0.2
node3:192.168.0.3
fence&qdisk:192.168.0.4
实验前准备
1.同步格服务器的时间（误差1分钟之内）
date -s
2.确认网关是否正确
route -n
3.开启ssh图形化
vim /etc/ssh/ssh_config
ForwardX11 yes
一、fence配置ssh key、hosts文件，发送到3台node上
ssh-keygen
ssh-copy-id -i ~/.ssh/id_rsa.pub 192.168.0.1
ssh-copy-id -i ~/.ssh/id_rsa.pub 192.168.0.2
ssh-copy-id -i ~/.ssh/id_rsa.pub 192.168.0.3
vim /etc/hosts
192.168.0.1 node1.test.com node1
192.168.0.2 node2.test.com node2
192.168.0.3 node3.test.com node3
192.168.0.4 fence.test.com fence
for i in 1 2 3 ;do scp /etc/hosts node$i:/etc/ ;done
二、安装fence所需的包，远程安装node上所需的包
for i in 1 2 3 ;do ssh node$i "yum groupinstall -y "Clustering"" ;done
for i in 1 2 3 ;do ssh node$i "yum install -y scsi-target-utils gfs2-utils lvm2-cluster mysql-server" ;done
yum groupinstall -y "Clustering"
yum install -y gnbd_serv iscsi-initiator-utils lvm2-cluster mysql-server
三、fence上导出两个iscsi共享
vim /etc/tgt/targets.conf
<target iqn.2011-12.com.test:fence.qdisk>
backing-store /dev/sda7
</target>
<target iqn.2011-12.com.test:fence.mysql>
backing-store /dev/vg0/lv0
</target>
/etc/init.d/tgtd start
chkconfig tgtd on
四、3台node导入iscsi共享，创建qdisk。
vim /etc/iscsi/initiatorname.iscsi
InitiatorName=iqn.2011-12.com.test:fence.qdisk
InitiatorName=iqn.2011-12.com.test:fence.mysql
/etc/init.d/iscsi start
iscsiadm -m discovery -t st -p 192.168.0.4
/etc/init.d/iscsi restart
mkqdisk -c /dev/sda -l myqdisk （在一台node上操作即可）
五、使用system-config-cluster配置cluster
system-config-cluster
1.New Configuration页
Choose a name for the cluster
clu_test
集群名称
Custom Configure Multicast
Address:225.225.225.225
广播地址
Use a Quorum Disk
Interval:
1投票周期1秒
TKO:10
10次投票没有成功
Votes:3
权重（大于n／2+1）
Minimum Score：1
最小分数1（只要有一分就可以工作）
Device：
设备文件名
Lable：myqdisk
设备标签名
Program:ping 192.168.0.1 -c2 -W1
判断是否应该投票的程序
Score:1
如果允许投票，每次投票的分数是1
Interval：2
间隔2秒投票一次
OK
2.Cluster Configuration页
Cluster Nodes
集群节点
Add a Cluster Node
Cluster Node Name:node1
OK
类似操作增加node2、node3、fence
Fence Devices
仲裁设备
Add a Fence Device
Global Network Block Device
Name:fd1
Servers:fence
OK
Cluster Nodes
节点关联仲裁
node1
Manage Fencing For This Node
Add a New Fence Level
Fence-Level-1
Add a New Fence to this Level
OK
类似操作为node2、node3，关联fence
Managed Resources
设置资源
Fallover Domains
Create a Failover Domain
Name for new Failover Domain:fd1
OK
Resources
Create a Resource
IP Address
192.168.0.5
浮动IP
OK
Create a Resource
Script
Name:mysql
File:/etc/init.d/mysqld
OK
Services
Create a Service
Name sv1
OK
Failover Domain:fd1
Add a Shared Resource to this service
加入IP资源、mysql资源
Close
保存退出
3.配置文件发送到node上
for i in 1 2 3 ;do scp /etc/cluster/cluster.conf node$i:/etc/cluster/cluster.conf ;done
六、配置mysql，启动集群
1.fence上启动相关服务
/etc/init.d/qdiskd start
/etc/init.d/cman start
gnbd_serv
2.在3台node上启动相关服务
/etc/init.d/qdiskd start
/etc/init.d/cman start
/etc/init.d/clvmd start
node1上用导入的mysql target创建一个lv0
3.检查node2、node3上是否同步了lv0，否则重启clvmd服务
4.格式化lv0（在一台node上操作即可）
mkfs.gfs2 -j3 -p lock_dlm -t clu_test:mysql /dev/vg0/lv0
5.修改mysqld脚本文件
vim /etc/init.d/mysqld
将
status mysqld >/dev/null >&1
if [ $? -eq 3 ] ; then
return 0
fi
加在 stop()函数的起始位置。
6.三台node上挂载lv0，启动rgmanager服务
mount /dev/vg0/lv0 /var/lib/mysql
/etc/init.d/rgmanager start