高可用pacemaker状态，db状态为stopped

admin

# pcs status
Cluster name: my_wsrep_cluster
Stack: 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
Last change: Thu Dec 28 18:05:11 2017 by root via cibadmin on controller1
3 nodes configured
, {8 a& l  Q3 l2 {' E  V1 resource configured
Online: [ controller1 controller2 controller3 ]
$ p5 C6 I) C- l- Q* eFull list of resources:
db (ocf::heartbeat:galera): Stopped
Failed Actions:
* db_start_0 on controller1 'not configured' (6): call=22, status=complete, exitreason='Galera must be configured as a multistate Master/Slave resource.',
9 q2 g- e: p2 Z* i+ L    last-rc-change='Thu Dec 28 18:05:11 2017', queued=1ms, exec=56ms
" w$ V. u4 {- C8 I$ W3 C6 F
Daemon Status:
  corosync: active/enabled
3 y. k: J6 G& v9 [  pacemaker: active/enabled
  pcsd: active/enabled
, j) g/ F) U+ t( c/ |5 _- P1 E
1 P! J; d* I9 c# C4 O0 ?8 }. N所有的状态都是active的，唯独只有db状态是stopped，为什么呢？
. S4 a8 t9 o  Y- A# }3 F6 z

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
/ Q" ]- @* W8 z0 F9 I# yum install mariadb-* -y
1 {2 E2 c- I: T# systemctl disable mariadb
% W. p- [1 m5 L& j" F. z# systemctl status mariadb
# pcs create resource DB ocf:heartbear:mysql
3 a. `. G& z( d2 U/ n7 p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
# 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 )
8 I. V3 j; B6 w4 b# pcs constraint location add server1 DB node1 1
, r2 c4 _! n6 [* W$ m9 X# pcs constraint location add server1 VIP1 node1 1
; h$ }9 E  }+ t7 I$ |7 B1 u# pcs constraint location add server2 FS node2 1
# pcs constraint location add server2 VIP node2 1
# pcs constraint location add server2 WEB node2 1
6 k5 C# p# q1 D- ?: j- V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'