I hope my upcoming question is specific enough, and you are able and willing to support :-) We have several openSUSE VMs in an ESX-Cluster (three ESX-Servers) with an attached iSCSI-SAN.

All of those Linux VMs are "single point of failure"-configured, which means in the case of a Web-Server: LAMP, storage, etc. everything on this machine. This was very simple and in case of a failure (in the last years: kernel panics or apache crashes) a simple reboot triggered by a script did it.

But the problem is: How to upgrade/maintain the w(eb-)application or the underlying OS without downtime? This wasn't really managable and i did this in the early morning ;) How can i achieve a "simple" High-Availability Cluster now?

I thought of: DRBD with heartbeat with 2 VMs. And for the storage a RDM (raw device mapped) LUN and change the read-write-permissions for both VMs.

Is this a good idea? Anyone has a better solution?