Hello,
I am planning to set up high availability like described in OpenNebula Front-end HA — OpenNebula 6.0.3 documentation, but I do not understand why an odd number of servers is required. What is the difference between on the whole only two servers and a fail of one of three servers with two servers remaining? Is an odd number of servers really absolutely necessary?
Regards
Christoph
Hi, put “ clustering why odd servers” into your google searchbar and you will see plenty of results regarding this topic. Then return back and ask if something is not clear to you 
Whatever happened to answering questions with an answer?
In simple terms, with two servers, if one goes down, the remaining server can’t be sure if it is just a local connectivity issue for itself, or if the other one has genuinely gone down - the same is true if the two servers just lose connectivity between themselves but are still functional, how do they know? With 3 servers, if one goes down or becomes split from the network, the other two can still see each other, so they know the problem is with the 3rd, so one of them should take over, they “form a consensus” on how to deal with it.