一次DRBD脑裂行为的模拟

drbd1为主，drbd2为辅；我个人觉得这个DRBD脑裂的行为，也应该是前期人为或是故障切换造成的，
如HA。 上次跟一朋友去一客户那里，他那就是属于使用HA做故障切换，最后不知道他们咋搞，在一台
机上把DRBD的服务给挂了，因为该服务器非常重要，他们对HA及DRBD架构不太熟，在一次HA切换测
试过程中出现了问题，在此模拟一下这个问题吧。

1、断开primary
down机或是断开网线
2、查看secondary机器的状态
[root@drbd2 ~]# drbdadm role fs
Secondary/Unknown

[root@drbd2 ~]# cat /proc/drbd
version: 8.3.11 (api:88/proto:86-96)
GIT-hash: 0de839cee13a4160eed6037c4bddd066645e23c5 build by root@drbd2.localdomain, 2011-07-08 11:10:20
#注意下drbd2的cs状态
1: cs:WFConnection ro:Secondary/Unknown ds:UpToDate/DUnknown C r-----
ns:567256 nr:20435468 dw:21002724 dr:169 al:229 bm:1248 lo:0 pe:0 ua:0 ap:0 ep:1 wo:b oos:0

将secondary配置成primary角色
[root@drbd2 ~]# drbdadm primary fs
[root@drbd2 ~]# drbdadm role fs
Primary/Unknown
[root@drbd2 ~]# cat /proc/drbd
version: 8.3.11 (api:88/proto:86-96)
GIT-hash: 0de839cee13a4160eed6037c4bddd066645e23c5 build by root@drbd2.localdomain, 2011-07-08 11:10:20

1: cs:WFConnection ro:Primary/Unknown ds:UpToDate/DUnknown C r-----
ns:567256 nr:20435468 dw:21002724 dr:169 al:229 bm:1248 lo:0 pe:0 ua:0 ap:0 ep:1 wo:b oos:0
#挂载
[root@drbd2 ~]# mount /dev/drbd1 /mnt/
[root@drbd2 ~]# cd /mnt/
[root@drbd2 mnt]# ll
total 102524
-rw-r--r-- 1 root root 104857600 Jul  8 12:35 100M
drwx------ 2 root root     16384 Jul  8 12:33 lost+found

#原来的primary机器好了，出现脑裂了。
[root@drbd1 ~]# tail -f /var/log/messages
Jul  8 13:14:01 localhost kernel: block drbd1: helper command: /sbin/drbdadm initial-split-brain minor-1 exit code 0 (0x0)
Jul  8 13:14:01 localhost kernel: block drbd1: Split-Brain detected but unresolved, dropping connection!
Jul  8 13:14:01 localhost kernel: block drbd1: helper command: /sbin/drbdadm split-brain minor-1
Jul  8 13:14:01 localhost kernel: block drbd1: helper command: /sbin/drbdadm split-brain minor-1 exit code 0 (0x0)
Jul  8 13:14:01 localhost kernel: block drbd1: conn( NetworkFailure -> Disconnecting )
Jul  8 13:14:01 localhost kernel: block drbd1: error receiving ReportState, l: 4!
Jul  8 13:14:01 localhost kernel: block drbd1: Connection closed
Jul  8 13:14:01 localhost kernel: block drbd1: conn( Disconnecting -> StandAlone )
Jul  8 13:14:01 localhost kernel: block drbd1: receiver terminated
Jul  8 13:14:01 localhost kernel: block drbd1: Terminating receiver thread

[root@drbd1 ~]# drbdadm role fs
Primary/Unknown

[root@drbd2 mnt]# drbdadm role fs
Primary/Unknown

#drbd1现在是standalone，这个时候，主跟辅是不会相互联系的。
[root@drbd1 ~]# cat /proc/drbd
version: 8.3.11 (api:88/proto:86-96)
GIT-hash: 0de839cee13a4160eed6037c4bddd066645e23c5 build by root@drbd1.localdomain, 2011-07-08 11:10:38

1: cs:StandAlone ro:Primary/Unknown ds:UpToDate/DUnknown   r-----
ns:20405516 nr:567256 dw:567376 dr:20405706 al:2 bm:1246 lo:0 pe:0 ua:0 ap:0 ep:1 wo:b oos:0

[root@drbd1 /]# service drbd status
drbd driver loaded OK; device status:
version: 8.3.11 (api:88/proto:86-96)
GIT-hash: 0de839cee13a4160eed6037c4bddd066645e23c5 build by root@drbd1.localdomain, 2011-07-08 11:10:38
m:res  cs          ro               ds                 p       mounted  fstype
1:fs   StandAlone  Primary/Unknown  UpToDate/DUnknown  r-----  ext3

这个时候，如果用户有尝试把drbd2的drbd服务重启的话，你就会发现根本无法起来！
[root@drbd2 /]# service drbd start
Starting DRBD resources: [ ]..........
***************************************************************
DRBD's startup script waits for the peer node(s) to appear.
- In case this node was already a degraded cluster before the
reboot the timeout is 120 seconds. [degr-wfc-timeout]
- If the peer was available before the reboot the timeout will
expire after 0 seconds. [wfc-timeout]
(These values are for resource 'fs'; 0 sec -> wait forever)
To abort waiting enter 'yes' [ -- ]:[  13]:[  15]:[  16]:[  18]:[  19]:[  20]:[  22]:

在drbd2处理方法：
[root@drbd2 /]# drbdadm disconnect fs
[root@drbd2 /]# drbdadm secondary fs
[root@drbd2 /]# drbdadm -- --discard-my-data fs
做完以上三步，你发现你仍然无法启动drbd2上的drbd服务；上次一客户我个人估计就是这个问题，把DRBD重启后，无法启动DRBD。

把他们DBA急的要死。呵呵

需要在drbd1上重连接资源：
[root@drbd1 ~]# drbdadm connect fs

再次启动drbd2上的drbd服务，成了。
[root@drbd2 /]# service drbd start
Starting DRBD resources: [ ].

再看看资源同步：
[root@drbd2 /]# cat /proc/drbd
version: 8.3.11 (api:88/proto:86-96)
GIT-hash: 0de839cee13a4160eed6037c4bddd066645e23c5 build by root@drbd2.localdomain, 2011-07-08 11:10:20

1: cs:SyncTarget ro:Secondary/Primary ds:Inconsistent/UpToDate C r-----
ns:0 nr:185532 dw:185532 dr:0 al:0 bm:15 lo:0 pe:0 ua:0 ap:0 ep:1 wo:b oos:299000
[======>.............] sync'ed: 39.5% (299000/484532)K
finish: 0:00:28 speed: 10,304 (10,304) want: 10,240 K/sec
大伙来拍砖吧！
补充：虽然是手工模拟但在故障切换时也会出一样的问题。
1、DRBD的资源只能在或主或辅的一台机器上挂载。
2、在做主辅的手工切换时的步骤：
a、先将原来挂载的东西进行卸载，这个时候你的应用会停，不建议手工切换主辅
b、将原来的主设置成辅 #drbdadm secondary resource_name
c、将原来的辅设置成主 #drbdadm primary resource_name
d、挂载资源