Re: configuration where 2 out of 3 physical servers can fail

[Ron Arts]

Mikael,

thanks for your detailed answer.

In my case the three VM's are located in 3 different datacenters about 10 miles apart. The chance of two failing at the same time can be assumed to be zero.

(I don't want to run circular replication because we as a supplier do not have access, and we cannot easily rely on MySQL Cluster-competent staff on site, so I want to keep the setup as simple as possible).

I'm leaning towards running with 3 replicas, since in my tests it worked well, Also shutdown will not happen at all, so the minor problem you mentioned does not bother me.

Your answer implies that if I colocate management daemons with data nodes, that if any single node fails, the cluster will go down if that happens to be the current arbitrator correct? Won't the other two management nodes negotiate a new arbitrator and thus keep the cluster up?

If not, I do have the option of running the management nodes on neighbouring-but-not-on-the-same-blade VMs. Would this alleviate the problem?

Thanks,
Ron