|
原帖由 cqubityj 于 2009-1-20 16:40 发表 ![]()
都放在同一个阵列上是否有些不大合适? 假设你用了3个fencing盘并且都放在存储A上,如果存储A停机了,那么等于说这个系统里一个fencing盘都没有了。如果这时候再出现split brain,不知VCS会如何处理?应该是两个节点都会重启吧。建议楼主测试一下(先关闭放fencing盘的存储,然后再把VCS的心跳线断了)。
我对VCS的fencing原理也不是很清楚。只知道在出现split brain的时候各个节点会争夺fencing盘,抢得多的就可以活下来,抢得少的就会重启。
原帖由 cqubityj 于 2009-1-20 11:13 发表 ![]()
请楼主详细说说VCS Coordinator Disks是怎么设置的?一台存储放两个,另一台存储放一个?还是......
Why do I need to place an arbitration point in a third site in case of a two-node stretch cluster?
A coordination point in a third site allows at least a single node cluster to continue operations in the event of a site failure in a two-node stretch cluster. The other site or node can still access the coordination point in the third site, apart from its local coordination point. This allows the surviving node to gain majority vote at the time of racing for coordination points. On the other hand, in the absence of a third coordination point, each site will, in all probability, have won the race for its local coordination point, and with one vote each, each node will commit suicide. |
|