Is it perceivable that one can experience data loss

in the failover of a VM from one node to another?

When VM1 is re-started on ESX2, it would be the in the same state as if you had killed the power on a physical server, ie. crashed state.  Depending on what was going on at the time of the crash, you may or may not lose data.  You can count on losing a few minutes in the reboot/cleanup that follows the HA failover though, as the whole process isn't instantaneous for VMs.

There seems to be misunderstandings with the HA/DRS features.  It can move running VMs around the cluster to balance load without interruption, thanks to vmotion.  However, once the physical host that's running a VM goes off-line, vmotion is not an option and all that HA can do is reboot the VM on another physical server and hope for the best.  Hope this clarifies it...

Agreed but really how different is this from the physically world. If a server crashes when it comes back up the server is in a crash consistent mode. Granted when an ESX server crashes (never seen it in my environment..knock on wood) you have a bunch of VMs that would restart in crash consistent mode.

The advantage in the VMWARE world is that the VMs actually get to restart on another piece of hardware. It is much more difficult to perform this in a traditional physical server environment without using clustering for each physical host and that gets expensive fast!

With VMWARE HA you receive High Availability for all your applications including the ones that don't support MS Clustering and there are alot of applications that do not support MSCS.

Following on from this discussion, is anyone aware of a Systems Management or Failover product that will be able to predict if a server is going to crash, and will be able to integrate with Vmotion to then move the necesarry VM's, therefore ensuring 0 (zero) downtime...?

I have looked into the many failover products on the market, and I have yet to find one which can integrate into VMware.

HP SIM offers similar features, and I would guess that Dell OpenManage does, too. Basically, if you go with a top-tier hardware vendor you can get predictive failure analysis features from their monitoring/management solutions...

HP & IBM both have products called VMM (Virtual Machine Management or something similar...) that are bolt-ons to SIM and Director, respectively. Each provides some degree of integration with VirtualCenter...