本文实验环境：
CentOS 6.5 x64 RHCS GFS
配置说明: 本文出自: http://koumm.blog.51cto.com
1. 通过Openfiler实现iscsi共享存储
2. 通过VMware ESXi5 虚拟fence实现fence功能。
3. 结合Centos 6.5 vmware-fence-soap实现RHCS fence设备功能。
4. 通过搭建RHCS实验环境测试GFS2功能。
RHEL/CentOS5.X版本RHCS配置相关参考:
IBM x3650M3+GFS+IPMI fence生产环境配置一例
http://koumm.blog.51cto.com/703525/1544971
Redhat 5.8 x64 RHCS Oracle 10gR2 HA实践配置（注：vmware-fence-soap）
http://koumm.blog.51cto.com/703525/1161791
一、准备基础环境
1. 网络环境准备
node01,node02节点
# cat /etc/hosts
192.168.0.181 node01.abc.com node01
192.168.0.182 node01.abc.com node02
2. 配置YUM安装源
(1) 挂载光盘ISO
# mount /dev/cdrom /mnt
(2) 配置YUM客户端
说明: 通过本地光盘做为yum安装源。
# vi /etc/yum.repos.d/rhel.repo
[rhel]
name=rhel6
baseurl=file:///mnt
enabled=1
gpgcheck=0
(3) openfiler iscsi存储配置
配置略,规划磁盘空间如下:
qdisk 100MB
data 150GB
(4) node01,node02挂载存储
# yum install iscsi-initiator-utils
# chkconfig iscsid on
# service iscsid start
# iscsiadm -m discovery -t sendtargets -p 192.168.0.187
192.168.0.187:3260,1 iqn.2006-01.com.openfiler:tsn.dea898a36535
# iscsiadm -m node -T iqn.2006-01.com.openfiler:tsn.dea898a36535 -p 192.168.0.187 –l
二、RHCS软件包的安装
1. 在node01上安装luci及RHCS软件包
1) node01（管理节点）安装RHCS软件包，luci是管理端软件包，只在管理端安装。
yum -y install luci cman odcluster ricci gfs2-utils rgmanager lvm2-cluster
2) node02 安装RHCS软件包
yum -y install cman odcluster ricci gfs2-utils rgmanager lvm2-cluster
3) node01, node02 更改个节点ricci用户密码
passwd ricci
4) 配置RHCS服务开机启动
chkconfig ricci on
chkconfig rgmanager on
chkconfig cman on
service ricci start
service rgmanager start
service cman start
#启动过程如下:
正在启动 oddjobd： [确定]
generating SSL certificates... done
Generating NSS database... done
启动 ricci： [确定]
Starting Cluster Service Manager: [确定]
Starting cluster:
Checking if cluster has been disabled at boot... [确定]
Checking Network Manager... [确定]
Global setup... [确定]
Loading kernel modules... [确定]
Mounting configfs... [确定]
Starting cman... xmlconfig cannot find /etc/cluster/cluster.conf
[失败]
Stopping cluster:
Leaving fence domain... [确定]
Stopping gfs_controld... [确定]
Stopping dlm_controld... [确定]
Stopping fenced... [确定]
Stopping cman... [确定]
Unloading kernel modules... [确定]
Unmounting configfs... [确定]
#
2. 在node01管理节点上安装启动luci服务
1) 启动luci服务
chkconfig luci on
service luci start
Adding following auto-detected host IDs (IP addresses/domain names), corresponding to `node01' address, to the configuration of self-managed certificate `/var/lib/luci/etc/cacert.config' (you can change them by editing `/var/lib/luci/etc/cacert.config', removing the generated certificate `/var/lib/luci/certs/host.pem' and restarting luci):
(none suitable found, you can still do it manually as mentioned above)
Generating a 2048 bit RSA private key
writing new private key to '/var/lib/luci/certs/host.pem'
Start luci... [确定]
Point your web browser to https://node01:8084 (or equivalent) to access luci
2) 配置管理地址, RHCS6版本采用root用户密码登录。
https://node01:8084
root/111111
三、RHCS集群配置
1. 添加集群
登录进管理界面，点击Manage Clusters --> Create 填入如下内容：
Cluster Name: gfs
NodeName Password RicciHostname Ricci Port
node01.abc.com 111111 node01.abc.com 11111
node02.abc.com 111111 node01.abc.com 11111
选中如下选项，然后提交
Use locally installed packages.
说明:这步会生成集群配置文件/etc/cluster/cluster.conf
2. 添加Fence Devices
说明:
RHCS要实现完整的集群功能，必须要实现fence功能。由于非物理服务器配置等条件限制，特使用VMware ESXi5.X的虚拟fence来实现fence设备的功能。
正是由于有了fence设备可以使用，才得以完整测试RHCS功能。
(1)登录进管理界面，点击cluster-> Fence Devices->
(2)选择"Add 选择VMware Fencing（SOAP Interface）
(3)Name "ESXi_fence"
(4)IP Address or Hostname "192.168.0.21"（ESXi地址）
(5)Login "root"
(6)Password "111111"
3. 节点绑定Fence设备
添加节点一fence
1) 点击node01.abc.com节点，Add Fence Method，这里填node01_fence;
2) 添加一个fence instance,选择"ESXi_fence" VMware Fencing（SOAP Interface）
3) VM NAME "kvm_node01"
4) VM UUID "564d6fbf-05fb-1dd1-fb66-7ea3c85dcfdf" 选中ssl
说明: VMNAME: 虚拟机名称，VM UUID: 虚拟机.vmx文件中"
uuid.location"值, 采用下面的字符串的格式。
# /usr/sbin/fence_vmware_soap -a 192.168.0.21 -z -l root -p 111111 -n kvm_node2 -o list
kvm_node2,564d4c42-e7fd-db62-3878-57f77df2475e
kvm_node1,564d6fbf-05fb-1dd1-fb66-7ea3c85dcfdf
添加节点二fence
1) 点击node02.abc.com节点，Add Fence Method，这里填node02_fence;
2) 添加一个fence instance,选择"ESXi_fence" VMware Fencing（SOAP Interface）
3) VM NAME "kvm_node02"
4) VM UUID "564d4c42-e7fd-db62-3878-57f77df2475e" 选中ssl
#手动测试fence功能示例:
# /usr/sbin/fence_vmware_soap -a 192.168.0.21 -z -l root -p 111111 -n kvm_node02 -o reboot
Status: ON
选项：
-o : list,status,reboot等参数
4. 添加Failover Domains配置
Name "gfs_failover"
Prioritized
Restricted
node01.abc.com 1
node02.abc.com 1
5. 配置GFS服务
(1) GFS服务配置
分别在node01,node02启动CLVM的集成cluster锁服务
lvmconf --enable-cluster
chkconfig clvmd on
service clvmd start
Activating VG(s): No volume groups found [ OK ]
(2) 在任意一节点对磁盘进行分区,划分出sdc1。然后格式化成gfs2.
node01节点上：
# pvcreate /dev/sdc1
Physical volume "/dev/sdc1" successfully created
# pvs
PV VG Fmt Attr PSize PFree
/dev/sda2 vg_node01 lvm2 a-- 39.51g 0
/dev/sdc1 lvm2 a-- 156.25g 156.25g
# vgcreate gfsvg /dev/sdc1
Clustered volume group "gfsvg" successfully created
# lvcreate -l +100%FREE -n data gfsvg
Logical volume "data" created
node02节点上：
# /etc/init.d/clvmd start
(3) 格式化GFS文件系统
node01节点上：
[root@node01 ~]# mkfs.gfs2 -p lock_dlm -t gfs:gfs2 -j 2 /dev/gfsvg/data
This will destroy any data on /dev/gfsvg/data.
It appears to contain: symbolic link to `../dm-2'
Are you sure you want to proceed? [y/n] y
Device: /dev/gfsvg/data
Blocksize: 4096
Device Size 156.25 GB (40958976 blocks)
Filesystem Size: 156.25 GB (40958975 blocks)
Journals: 2
Resource Groups: 625
Locking Protocol: "lock_dlm"
Lock Table: "gfs:gfs2"
UUID: e28655c6-29e6-b813-138f-0b22d3b15321
说明:
gfs:gfs2这个gfs就是集群的名字，gfs2是定义的名字，相当于标签。
-j是指定挂载这个文件系统的主机个数，不指定默认为1即为管理节点的。
这里实验有两个节点
6. 挂载GFS文件系统
node01,node02 上创建GFS挂载点
# mkdir /vmdata
(1)node01,node02手动挂载测试，挂载成功后，创建文件测试集群文件系统情况。
# mount.gfs2 /dev/gfsvg/data /vmdata
(2)配置开机自动挂载
# vi /etc/fstab
/dev/gfsvg/data /vmdata gfs2 defaults 0 0
[root@node01 vmdata]# df -h
Filesystem Size Used Avail Use% Mounted on
/dev/mapper/vg_node01-lv_root 36G 3.8G 30G 12% /
tmpfs 1.9G 32M 1.9G 2% /dev/shm
/dev/sda1 485M 39M 421M 9% /boot
/dev/gfsvg/data 157G 259M 156G 1% /vmdata
7. 配置表决磁盘
说明:
#表决磁盘是共享磁盘，无需要太大，本例采用/dev/sdc1 100MB来进行创建。
[root@node01 ~]# fdisk -l
Disk /dev/sdb: 134 MB, 134217728 bytes
5 heads, 52 sectors/track, 1008 cylinders
Units = cylinders of 260 * 512 = 133120 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x80cdfae9
Device Boot Start End Blocks Id System
/dev/sdb1 1 1008 131014 83 Linux
(1) 创建表决磁盘
[root@node01 ~]# mkqdisk -c /dev/sdb1 -l myqdisk
mkqdisk v0.6.0
Writing new quorum disk label 'myqdisk' to /dev/sdc1.
WARNING: About to destroy all data on /dev/sdc1; proceed [N/y] ? y
Initializing status block for node 1...
Initializing status block for node 2...
Initializing status block for node 3...
Initializing status block for node 4...
Initializing status block for node 5...
Initializing status block for node 6...
Initializing status block for node 7...
Initializing status block for node 8...
Initializing status block for node 9...
Initializing status block for node 10...
Initializing status block for node 11...
Initializing status block for node 12...
Initializing status block for node 13...
Initializing status block for node 14...
Initializing status block for node 15...
Initializing status block for node 16...
(2) 查看表决磁盘信息
[root@node01 ~]# mkqdisk -L
mkqdisk v3.0.12.1
/dev/block/8:17:
/dev/disk/by-id/scsi-14f504e46494c455242553273306c2d4b72697a2d544e6b4f-part1:
/dev/disk/by-path/ip-192.168.0.187:3260-iscsi-iqn.2006-01.com.openfiler:tsn.dea898a36535-lun-0-part1:
/dev/sdb1:
Magic: eb7a62c2
Label: myqdisk
Created: Thu Jan 1 23:42:00 2015
Host: node02.abc.com
Kernel Sector Size: 512
Recorded Sector Size: 512
(3) 配置表决磁盘qdisk
# 进入管理界面Manage Clusters --> gfs --> Configure --> QDisk
Device : /dev/sdc1
Path to program : ping -c3 -t2 192.168.0.253
Interval : 3
Score : 2
TKO : 10
Minimum Score : 1
# 点击apply
(4) 启动qdisk服务
chkconfig qdiskd on
service qdiskd start
clustat -l
[root@node01 ~]# clustat -l
Cluster Status for gfs @ Thu Jan 1 23:50:53 2015
Member Status: Quorate
Member Name ID Status
------ ---- ---- ------
node01.abc.com 1 Online, Local
node02.abc.com 2 Online
/dev/sdb1 0 Online, Quorum Disk
[root@node01 ~]#
8. 测试GFS
1）node02节点上执行
# echo c > /proc/sysrq-trigger
2）node01节点上查看日志记录
# tail -f /var/log/messages
Jan 2 01:37:47 node01 ricci: startup succeeded
Jan 2 01:37:47 node01 rgmanager[2196]: I am node #1
Jan 2 01:37:47 node01 rgmanager[2196]: Resource Group Manager Starting
Jan 2 01:37:47 node01 rgmanager[2196]: Loading Service Data
Jan 2 01:37:49 node01 rgmanager[2196]: Initializing Services
Jan 2 01:37:49 node01 rgmanager[2196]: Services Initialized
Jan 2 01:37:49 node01 rgmanager[2196]: State change: Local UP
Jan 2 01:37:49 node01 rgmanager[2196]: State change: node02.abc.com UP
Jan 2 01:37:52 node01 polkitd[3125]: started daemon version 0.96 using authority implementation `local' version `0.96'
Jan 2 01:37:52 node01 rtkit-daemon[3131]: Sucessfully made thread 3129 of process 3129 (/usr/bin/pulseaudio) owned by '42' high priority at nice level -11.
Jan 2 01:40:52 node01 qdiskd[1430]: Assuming master role
Jan 2 01:40:53 node01 qdiskd[1430]: Writing eviction notice for node 2
Jan 2 01:40:54 node01 qdiskd[1430]: Node 2 evicted
Jan 2 01:40:55 node01 corosync[1378]: [TOTEM ] A processor failed, forming new configuration.
Jan 2 01:40:57 node01 corosync[1378]: [QUORUM] Members[1]: 1
Jan 2 01:40:57 node01 corosync[1378]: [TOTEM ] A processor joined or left the membership and a new membership was formed.
Jan 2 01:40:57 node01 kernel: dlm: closing connection to node 2
Jan 2 01:40:57 node01 corosync[1378]: [CPG ] chosen downlist: sender r(0) ip(192.168.0.181) ; members(old:2 left:1)
Jan 2 01:40:57 node01 corosync[1378]: [MAIN ] Completed service synchronization, ready to provide service.
Jan 2 01:40:57 node01 kernel: GFS2: fsid=gfs:gfs2.1: jid=0: Trying to acquire journal lock...
Jan 2 01:40:57 node01 fenced[1522]: fencing node node02.abc.com
Jan 2 01:40:57 node01 rgmanager[2196]: State change: node02.abc.com DOWN
Jan 2 01:41:11 node01 fenced[1522]: fence node02.abc.com success
Jan 2 01:41:12 node01 kernel: GFS2: fsid=gfs:gfs2.1: jid=0: Looking at journal...
Jan 2 01:41:12 node01 kernel: GFS2: fsid=gfs:gfs2.1: jid=0: Done
Jan 2 01:41:30 node01 corosync[1378]: [TOTEM ] A processor joined or left the membership and a new membership was formed.
Jan 2 01:41:30 node01 corosync[1378]: [QUORUM] Members[2]: 1 2
Jan 2 01:41:30 node01 corosync[1378]: [QUORUM] Members[2]: 1 2
Jan 2 01:41:30 node01 corosync[1378]: [CPG ] chosen downlist: sender r(0) ip(192.168.0.181) ; members(old:1 left:0)
Jan 2 01:41:30 node01 corosync[1378]: [MAIN ] Completed service synchronization, ready to provide service.
Jan 2 01:41:38 node01 qdiskd[1430]: Node 2 shutdown
Jan 2 01:41:50 node01 kernel: dlm: got connection from 2
Jan 2 01:41:59 node01 rgmanager[2196]: State change: node02.abc.com UP
说明：fence功能正常，期间GFS文件系统正常。
9. 配置文件
cat /etc/cluster/cluster.conf
配置小结：
1, 如果采用FC存储，就不会出现IP SAN因网络出现的问题。
2, IP SAN配置时，最好采用专用的网卡与网段。在安装成功后，因配置KVM虚拟环境，调整网卡为桥接接口，虽然网络是通的，但是无法启动GFS, 最终在国外网站上找到一句话，才解释了原因，最终连接IPSAN网卡接口为固定接口，GFS环境恢复正常。
https://www.mail-archive.com/[email protected]/msg03800.html
You'll need to check the routing of the interfaces. The most common cause of this sort of error is having two interfaces on the same physical (or internal) network.
3, 实际环境可以采用真实fence设备来带替fence_vmware_soap，实现fence功能。