1. 3个节点一个集群
2. 一个fail domain 节点1 优先级 3
节点2 优先级 1
节点3 优先级 2
Feb 25 22:18:21 app3 ricci[6354]: Executing ‘/usr/libexec/ricci/ricci-worker -f /var/lib/ricci/queue/909647949‘
Feb 25 22:18:25 app3 ricci[6363]: Executing ‘/usr/libexec/ricci/ricci-worker -f /var/lib/ricci/queue/1138168200‘
Feb 25 22:19:06 app3 clurgmgrd[4636]: <notice> Relocating service:appvip to better node app2
Feb 25 22:19:06 app3 clurgmgrd[4636]: <notice> Stopping service service:appvip
Feb 25 22:19:06 app3 avahi-daemon[4557]: Withdrawing address record for 172.16.8.55 on eth1.
Feb 25 22:19:16 app3 clurgmgrd[4636]: <notice> Service service:appvip is stopped
Feb 25 22:19:51 app3 clurgmgrd[4636]: <notice> Member 2 shutting down
Feb 25 22:19:57 app3 clurgmgrd[4636]: <notice> Starting stopped service service:appvip
Feb 25 22:19:59 app3 avahi-daemon[4557]: Registering new address record for 172.16.8.55 on eth1.
Feb 25 22:20:00 app3 clurgmgrd[4636]: <notice> Service service:appvip started
之前vip启动在节点3上，关闭节点3的rgmanager 服务，vip 启动到了节点2上，启动节点3的rgmanager，vip重新回到了节点3
clusvcadm relocate定位到app2上，关闭vip服务，enable，自动vip从app2 上，回到了app3上 Member Name ID Status
------ ---- ---- ------
app3 1 Online, rgmanager
app2 2 Online, Local, rgmanager
app1 3 Online, rgmanager
目前怀疑是member 顺序的问题，优先级似乎不觉得启动在哪个节点上，只是fail了，启动到哪个节点
结论： stop 再enable 服务 他会到集群member 1的服务器上
restart 服务，关闭的是在哪里，启动还是在那台服务器上
把节点1剔除fail domain 后测试，限制在fail domain 里面切换vip，这样就只会在节点2 和节点3之间切换，默认启动是在节点2上了，关节点2 rgmanager服务，自动漂移到节点3，手动relocate 到节点1报错，禁止非fail domain漂移，开启节点2 rgmanager服务，自动回收vip服务到节点2，stop vip服务，enable 后，启动还是在member 1上，也就是节点3
本文出自 “骅骝向北越鸟归南” 博客，请务必保留此出处http://848436.blog.51cto.com/838436/1363311
原文：http://848436.blog.51cto.com/838436/1363311