1753300 會員
7146 線上
108792 解決方案
發表新文章

cluster votes一問

 
watermelonyu
教授

cluster votes一問

two member cluster recommended 裡面寫著

each of which has 1 node vote

and quorum disk with 1 vote

Each member has value of 3 for expected votes

我有個疑問

當node 1 與node 2互相找不到對方時

但是node 1 and node2卻都還可以看到quorum disk

這時候 1+1>=,所以兩個nodes 應該都還會活著

請問這樣會導致cluster partition嗎
5則回覆 5
watermelonyu
教授

cluster votes一問

Hi Joey,



A network partition encountered as your description.

And it should be rebooting someone cluster member automatically or crash

one member to get the ICS connect communication error.



Hope it help to you.



Best regards,

Richard.
watermelonyu
教授

cluster votes一問

可以從哪裡判斷哪一台會get 到ICS error

所以我可以知道哪一台會reboot or crash



watermelonyu
教授

cluster votes一問

Hi Joey,



When communication failures occur, it is the connection manager's job (this is a thread who includes in the process "kernel_idle". Check it from command "ps -em -O THREAD|grep cnx) to make sure that only one cluster is formed (or maintained).



The connection manager will detect the problem and reconfigure the cluster to remove any mamber without complete communications connectivity to the other cluster members. Any member that cann't fully communicate with the other cluster members will be prevented from joining the cluster, and in many cases, it will panic.



Show the possible panic strings while cluster partition encounter:

"CNX MGR: partition action"

"CNX MGR: this node removed from cluster"

"CNX MGR: rcnx_status: restart requested to resynchronize with cluster wirh quorum"

"CNX QDISK: configuration error. Qdisk in use by cluster of different name."

"CNX QDISK: Yielding to foreign owner with (or without) quorum."



So that is possible to get both crash on both cluster member. And it needs to build two ICS from netrain configuration or two Memory Channel for network redundant.



Best regards,

Richard.
watermelonyu
教授

cluster votes一問

再請問一下,

用兩個memory channel ,是做redundant 還是可以做load balance or netrain
watermelonyu
教授

cluster votes一問

Hi Joey,



Both netrain (Fast ethernet interfaces) or two MCs (Memory Channels) building an ICS in cluster configuration that's always only one active, it can't to do load balance.



Best regards,

Richard.