oosefs分布式文件系统是一个易用的系统,但其只有在Pro版中提供了master的高可用方案,免费版master只能单机运行,存在单点故障的隐患。
本文结合网上的相关资料,介绍通过drbd keepalived来实现mfsmaster高可用的方案。
环境:
CentOS 6
Master-primary IP: 172.18.18.201 (主机名test01)
Master-secondary IP: 172.18.18.202 (test02)
Mfschunkserver IP:172.18.18.203 (test03)
Master 虚拟IP:172.18.18.204
一、安装操作系统
安装CentOS 6系统。需注意:在安装mfsmaster服务器时,一定要划分出一个独立的分区给drbd使用,我这里用的是/dev/sda3,大小500MB。
二、安装drbd
2.1 环境准备
因drbd编辑安装时,需要用到CentOS内核代码,因此需下载对应内核的源码包到本地安装,不要用yum安装。
rpm -ikernel-devel-2.6.32-504.el6.x86_64.rpm
安装编译时需要的环境:
yum -y install gcc flex perl
2.2 drbd编辑安装
tar zxvf drbd-8.4.2.tar.gz
cd drbd-8.4.2
./configure --prefix=/usr/local/drbd–with-km
makeKDIR=/usr/src/kernels/2.6.32-504.el6.x86_64
–说明:这是实际内核源码路径,根据实际情况设定
make install
mkdir -p /usr/local/drbd/var/run/drbd
cp /usr/local/drbd/etc/rc.d/init.d/drbd/etc/rc.d/init.d
chkconfig --add drbd
chkconfig drbd on
安装DRBD模块:
cd /root/data/drbd-8.4.2/drbd
make clean
makeKDIR=/usr/src/kernels/2.6.32-504.el6.x86_64
cp drbd.ko/lib/modules/2.6.32-504.el6.x86_64/kernel/lib/
–说明:内核版本要用uname -r查一下
modprobe drbd
查看模块是否加载成功
lsmod |grep drbd
drbd 299688 0
libcrc32c 1246 1 drbd
2.3 drbd配置
vi/usr/local/drbd/etc/drbd.d/global_common.conf
global {
usage-count yes;
}
common {
net {
代码语言:javascript复制protocol C;
}
}
vi /usr/local/drbd/etc/drbd.d/r0.res
resource r0 {
on test01 {
代码语言:javascript复制 device /dev/drbd1;
disk /dev/sda3;
address 172.18.18.201:7788;
meta-disk internal;
}
on test02 {
代码语言:javascript复制 device /dev/drbd1;
disk /dev/sda3;
address 172.18.18.202:7788;
meta-disk internal;
}
}
创建drbd资源:
dd if=/dev/zero of=/dev/sda3 bs=1Mcount=1 --一定要执行这个否则下一步会报错
drbdadm create-md r0
drbdadm up r0
以上操作在secondary (test02)上也执行一遍。
2.4 设置Primary Node
drbdadm primary --force r0
查看drbd状态:
cat /proc/drbd
1:cs:SyncSource ro:Primary/Secondary ds:UpToDate/Inconsistent C r—n-
ns:108008 nr:0 dw:0 dr:109208 al:0 bm:6 lo:0 pe:1 ua:1 ap:0 ep:1 wo:foos:404428
代码语言:javascript复制 [===>................]sync'ed: 21.6% (404428/511948)K
finish:0:00:33 speed: 11,944 (11,944) K/sec
2.5 创建drbd文件系统
在primary上:
mkfs.ext4 /dev/drbd1
mkdir drbddata
mount /dev/drbd1 /drbddata
2.6 drbd同步测试
在primary (test01)上,往/drbddata目录中写入测试文件后,执行:
umount /dev/drbd1
drbdadm secondary r0
然后在secondary (test02)上,执行:
mkdir drbddata
drbdadm primary r0
mount /dev/drbd1 /drbddata
此时,在/drbddata目录应能看到刚写入的测试文件!
三、安装mfs
3.1 环境准备
安装libpcap
yum –y install libpcap
3.2 安装mfsmaster
tar zxvfmoosefs-packages-linux-3.0.77.tar.gz
rpm -imoosefs-master-3.0.77-1.rhsysv.x86_64.rpm
3.3 配置mfsmaster
cd /etc/mfs
vi mfsmaster.cfg
确保修改:DATA_PATH = /drbddata
其它按默认参数设置。
以上操作在secondary (test02)上也执行一遍,完成两台服务器mfsmaster的安装。
3.4 mfsmaster启动测试
在primary上启动测试:
drbdadm primary r0
mount /dev/drbd1 /drbddata
chown mfs:mfs –R /drbddata
mfsmaster start
查看日志看mfsmaster是否启动成功。
在secondary上启动测试:
先将primary上服务停止:
mfsmaster stop
umount /drbddata
drbdadm secondary r0
再在secondary上执行:
drbdadm primary r0
mount /dev/drbd1 /drbddata
chown mfs:mfs –R /drbddata
mfsmaster start
测试成功后,进行下一步的安装。
四、安装keepalived
4.1 安装keepalived
tar zxvf keepalived-1.2.22.tar.gz
cd keepalived-1.2.22
./configure --prefix=/usr/local/keepalived --disable-fwmark
make
make install
cp /usr/local/keepalived/sbin/keepalived/usr/sbin/
cp /usr/local/keepalived/etc/sysconfig/keepalived/etc/sysconfig/
cp/usr/local/keepalived/etc/rc.d/init.d/keepalived /etc/init.d/
chkconfig --add keepalived
chkconfig keepalived on
4.2 配置keepalived
mkdir -p /etc/keepalived
cp/usr/local/keepalived/etc/keepalived/keepalived.conf /etc/keepalived/
vi /etc/keepalived/keepalived.conf
global_defs {
notification_email {
shuyb@sina.com
}
notification_email_from shuyb@sina.com
smtp_server 192.168.200.1
smtp_connect_timeout 30
router_id test01
vrrp_skip_check_adv_addr
vrrp_strict
vrrp_garp_interval 0
vrrp_gna_interval 0
}
vrrp_script check_drbd {
script “/etc/keepalived/check_drbd.sh”
interval 15
}
vrrp_instance mfs {
代码语言:javascript复制state MASTER
interface p4p1
virtual_router_id 51
priority 100
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.18.18.204
}
track_script {
check_drbd
}
}
在/etc/keepalived下新增一个文件:
vi check_drbd.sh
#!/bin/bash
A=ps -C mfsmaster --no-header |wc -l
if [ $A -eq 0 ]; then
umount /dev/drbd1
drbdadm secondary r0
killall keepalived
fi
chmod x check_drbd.sh
在secondary上安装步骤同上,配置文件如下:
vi /etc/keepalived/keepalived.conf
global_defs {
notification_email {
shuyb@sina.com
}
notification_email_from shuyb@sina.com
smtp_server 192.168.200.1
smtp_connect_timeout 30
router_id test01
vrrp_skip_check_adv_addr
vrrp_strict
vrrp_garp_interval 0
vrrp_gna_interval 0
}
vrrp_instance mfs {
代码语言:javascript复制state BACKUP
interface p4p1
virtual_router_id 51
priority 90
advert_int 1
authentication {
auth_type PASS
auth_pass 1111
}
virtual_ipaddress {
172.18.18.204
}
notify_master /etc/keepalived/master.sh
notify_backup /etc/keepalived/backup.sh
}
在/etc/keepalived下新增两个文件:
vi backup.sh
#!/bin/bash
mfsmaster stop
umount /dev/drbd1
drbdadm secondary r0
vi master.sh
#!/bin/bash
drbdadm primary r0
mount /dev/drbd1 /drbddata
mfsmaster start
chmod x backup.sh master.sh
至此,安装配置完成。需要注意的是,要将服务器的防火墙关闭,或保证mfs需要用到的9420~9425端口、drbd用到的7788端口等打开!
五、安装mfschunkserver和mfs客户端
5.1在chunkserver(test03)上
yum install libpcap
rpm -imoosefs-chunkserver-3.0.77-1.rhsysv.x86_64.rpm
vi /etc/mfs/mfsmetalogger.cfg
修改:
MASTER_HOST = 172.18.18.204
vi /etc/mfs/mfshdd.cfg
增加:
/home
(注:/home是该服务器上给mfschunk用的目录,mount到一个独立的逻辑卷上)
chown -R mfs:mfs /home
5.2 在mfs客户端
yum install fuse-libs
rpm -imoosefs-client-3.0.77-1.rhsysv.x86_64.rpm
mkdir -p /mfs
六、启动及切换测试
好了,现在可以进行测试了!
1. 先把mfsmaster在primary上启动起来,步骤同上
2. 在primary上启动keepalived
/etc/init.d/keepalived start
查看虚拟ip是否绑定:
ip addr
3.在secondary上启动keepalived
/etc/init.d/keepalived start
- 启动mfschunkserver
mfschunkserver start
查看/var/log/message看与Master是否连接成功
- 客户端连接
mfsmount /mfs –H 172.18.18.204
df --查看是否挂载成功
- 停止primary上的mfsmaster
mfsmaster stop
- 查看secondary、chunkserver、client是否正常
测试完成。完成后,需手工将服务重新切回到primary上。
ps:当出现问题调转到backup后,如果主的正常了,我们手动切换必须遵守以下步骤
停掉mfsmaster服务
mfsmaster stop
代码语言:javascript复制 umount /dev/drbd1
drbdadm secondary mfs
主机:
1.drbdadm primary mfs
代码语言:javascript复制 2.mount /dev/drbd1 /data1/drbd
3.mfsmaster start
4.keepalived start
还有要提到的就是可能会出现脑裂,这一般时操作不按顺序,也可能有其他原因,下面是网上大神写的,
复制代码 1,正常情况下状态: [root@drbd1 ~]# cat /proc/drbd version: 8.3.8 (api:88/proto:86-94) : 299AFE04D7AFD98B3CA0AF9 0: cs:Connected ro:Primary/Secondary ds:UpToDate/UpToDate C r---- ns:2144476 nr:0 dw:36468 dr:2115769 al:14 bm:129 lo:0 pe:0 ua:0 ap:0 ep:1 wo:b oos:0
[root@drbd2 ~]# cat /proc/drbd
version: 8.3.8 (api:88/proto:86-94)
srcversion: 299AFE04D7AFD98B3CA0AF9
0: cs:Connected ro:Secondary/Primary ds:UpToDate/UpToDate C r----
代码语言:javascript复制ns:0 nr:2141684 dw:2141684 dr:0 al:0 bm:130 lo:0 pe:0 ua:0 ap:0 ep:1 wo:b oos:0
2,drbd1故障后
drbd1状态:
[root@drbd1 ~]# cat /proc/drbd
version: 8.3.8 (api:88/proto:86-94)
srcversion: 299AFE04D7AFD98B3CA0AF9
0: cs:StandAlone ro:Primary/Unknown ds:UpToDate/DUnknown r----
代码语言:javascript复制ns:4 nr:102664 dw:102668 dr:157 al:1 bm:8 lo:0 pe:0 ua:0 ap:0 ep:1 wo:b oos:0
drbd2的状态:
[root@drbd2 ~]# cat /proc/drbd
version: 8.3.8 (api:88/proto:86-94)
srcversion: 299AFE04D7AFD98B3CA0AF9
0: cs:WFConnection ro:Secondary/Unknown ds:UpToDate/DUnknown C r----
代码语言:javascript复制ns:0 nr:2141684 dw:2141684 dr:0 al:0 bm:130 lo:0 pe:0 ua:0 ap:0 ep:1 wo:b oos:0
3,处理方法:
a,将secondary配置成primary角色
[root@drbd2 ~]# drbdsetup /dev/drbd0 primary -o
[root@drbd2 ~]# cat /proc/drbd
version: 8.3.8 (api:88/proto:86-94)
srcversion: 299AFE04D7AFD98B3CA0AF9
0: cs:WFConnection ro:Primary/Unknown ds:UpToDate/Outdated C r----
代码语言:javascript复制ns:0 nr:2141684 dw:2141684 dr:0 al:0 bm:130 lo:0 pe:0 ua:0 ap:0 ep:1 wo:b oos:0
挂载:
[root@drbd2 /]# mount /dev/drbd0 /data1
[root@drbd2 data1]# ll
total 10272
-rw-r–r-- 1 root root 10485760 Feb 13 11:26 aa.img
drwx------ 2 root root 16384 Feb 13 11:25 lost found
这个时候drbd2开始提供服务,开始写数据
drbd1主恢复正常后:
[root@drbd1 ~]# cat /proc/drbd
version: 8.3.8 (api:88/proto:86-94)
srcversion: 299AFE04D7AFD98B3CA0AF9
0: cs:StandAlone ro:Primary/Unknown ds:UpToDate/DUnknown r----
代码语言:javascript复制ns:2144476 nr:0 dw:36484 dr:2115769 al:14 bm:129 lo:0 pe:0 ua:0 ap:0 ep:1 wo:b oos:8
drbd1状态是:StandAlone,此时,drbd1是不会和drbd2互相联系的
我们来查看下日志:
[root@drbd1 ~]# tailf /var/log/messages
Feb 13 16:14:27 drbd1 kernel: block drbd0: helper command: /sbin/drbdadm split-brain minor-0
Feb 13 16:14:27 drbd1 kernel: block drbd0: helper command: /sbin/drbdadm split-brain minor-0 exit code 0 (0x0)
Feb 13 16:14:27 drbd1 kernel: block drbd0: conn( WFReportParams -> Disconnecting )
Feb 13 16:14:27 drbd1 kernel: block drbd0: error receiving ReportState, l: 4!
Feb 13 16:14:27 drbd1 kernel: block drbd0: asender terminated
Feb 13 16:14:27 drbd1 kernel: block drbd0: Terminating drbd0_asender
Feb 13 16:14:27 drbd1 kernel: block drbd0: Connection closed
Feb 13 16:14:27 drbd1 kernel: block drbd0: conn( Disconnecting -> StandAlone )
Feb 13 16:14:27 drbd1 kernel: block drbd0: receiver terminated
Feb 13 16:14:27 drbd1 kernel: block drbd0: Terminating drbd0_receiver
脑裂出现!
解决方法:
1>,我们需要将现在的drbd1角色修改为secondary
[root@drbd1 ~]# drbdadm secondary r0
[root@drbd1 ~]# drbdadm – --discard-my-data connect r0 ##该命令告诉drbd,secondary上的数据不正确,以primary上的数据为准。
2>,我们还需要在drbd2上执行下面操作
[root@drbd2 /]# drbdadm connect r0
这样drbd1就能和drbd2开始连接上了,并且保证数据不会丢失:
[root@drbd1 ~]# cat /proc/drbd
version: 8.3.8 (api:88/proto:86-94)
srcversion: 299AFE04D7AFD98B3CA0AF9
0: cs:Connected ro:Secondary/Primary ds:UpToDate/UpToDate C r----
代码语言:javascript复制ns:0 nr:20592 dw:20592 dr:0 al:0 bm:4 lo:0 pe:0 ua:0 ap:0 ep:1 wo:b oos:0