高可用pacemaker状态，db状态为stopped

admin

# pcs status
Cluster name: my_wsrep_cluster
; X: }, `4 t/ N1 a. L5 qStack: corosync
Current DC: controller1 (version 1.1.16-12.el7_4.5-94ff4df) - partition with quorum
Last updated: Thu Dec 28 18:24:33 2017
' c( C& @% H6 M1 s" n2 G( [Last change: Thu Dec 28 18:05:11 2017 by root via cibadmin on controller1
3 nodes configured
, e3 D# |: B- G0 e) G$ x1 resource configured
/ Z. f7 H( W! o0 |0 I/ S' pOnline: [ controller1 controller2 controller3 ]
Full list of resources:
db (ocf::heartbeat:galera): Stopped
9 R  G, h; n; V0 J4 b7 CFailed Actions:
9 s" w. p8 x" j$ p+ Q: @" N* db_start_0 on controller1 'not configured' (6): call=22, status=complete, exitreason='Galera must be configured as a multistate Master/Slave resource.',
    last-rc-change='Thu Dec 28 18:05:11 2017', queued=1ms, exec=56ms
' u6 v$ ^. y: ^, L% t: L2 v! N1 N' Y
Daemon Status:
$ ~% J& t% a) O$ c  corosync: active/enabled
' m6 }* }5 D2 _: w  z) P' Q  pacemaker: active/enabled
3 A+ S( `) Y* C. S- w; C  pcsd: active/enabled

所有的状态都是active的，唯独只有db状态是stopped，为什么呢？
8 V8 s9 ?% M1 t1 }

admin

才疏学浅，未能解决问题，遗憾。不知哪儿出问题了，好尴尬。

admin

配置一个mariadb服务
在所有节点上执行
# yum install mariadb-* -y
# systemctl disable mariadb
: q2 w3 p2 E8 x' b# systemctl status mariadb
. J- I- v$ B: E3 p# pcs create resource DB ocf:heartbear:mysql
3 u6 ?( ?! S- ]4 l配置一个互为备份的双机环境，一个节点作为web，一个节点作为db，各自有自己的浮动IP，出现故障的时候互相切换
# pcs resource create VIP1 ocf:heartbeat:IPaddr2 ip=192.168.122.171 cidr_netmask=24 nic=eth0
5 T* c! k* c( m4 d& `& d5 c# pcs resource group add MyGroup1 VIP1 DB(因为前面已经把DB添加到MyGroup组，所以需要先用pcs resource disable WEB; pcs resource diable DB;pcs resource group remove MyGroup DB; pcs resource enable WEB;才可以添加DB到MyGroup1 )
# pcs constraint location add server1 DB node1 1
1 z* N7 Q- y% g5 q" c' }+ y$ @# pcs constraint location add server1 VIP1 node1 1
, j- g+ U( j5 r8 o# pcs constraint location add server2 FS node2 1
# pcs constraint location add server2 VIP node2 1
3 i. [% [9 c" D6 T" ^# Z# pcs constraint location add server2 WEB node2 1
  r3 s& u5 ~) j% R执行上述命令后，DB和VIP1将启动在node1上而剩下的FS VIP WEB将启动在node2上

admin

突然想起来好像用的是pike版本的源，这个版本有可能有问题。还需要验证，想着回退到octa版本，试试。

admin

controller_start_0 on controller1 'unknown error'