Hi,

I am looking at creating a MySQL cluster as a back end to a site I am in the process of setting up. The hosting company I am with prefers virtual machines, in particular VMWare virtual machines. I just read the mysql-cluster faq's found here and saw that it is not recommended to run a mysql cluster on virtual machines. Is this true? or can you in fact run a mysql cluster on something other than physical machines? Has anyone tried to run a mysql cluster on VMs before?

SQL on a VMWare cluster is not necessarily required IMO. If you are using a well designed VMWare cluster you will have a cluster of hosts and duplicated / redundant back end storage. So in essence you are clustering on a cluster.

If you do have to different Virtual Machines ensure that they are on a separate VMWare cluster, storage and node else there is no point in having two virtual machines clustered on the same hardware.

In a nut shell this depends on your providers setup.

i dont see the point either, a virtual machine eats up resources that would otherwise be available to your SQL server, if you are running multiple instances of vmware on the same machine that would slow the server to a crawl, if you are running vmware on multiple machines then you should just cluster those instead
that being said a virtual machine should behave like any other machine and can probably be clustered the same way via the vmware virtual network, however unless there is a specific reason to use virtual machines it seems like an unnecessary step and a waste of resources as now the server has to process the emulated machine as well as the sql

Thank you for your replies. I know that it's a bad idea performance wise to have your MySQL Cluster running on VM's. I just wanted to know whether it was possible or not (I don't know if all of the VM's would be on the same physical machine or not. Have to ask hosting company). It seems that the problem of running each node on a separate VM is that the extra latency of the VM layer causes the nodes to think that some of the other nodes have crashed. (Thanks to Monty for the answer here).

This would mean that it would be possible, however, if the cluster started having problems and behaving oddly, it would have to have it's timeout values reconfigured.