Life migrating a VM from a failed host to another is like riding a dead horse, neither will work.
Option for such cases where downtime has to be minimized could be FT (Fault Tolerance), or some kind of guest OS clustering.
As above, if a host suddenly dies then the VMs running on it dies also. Unless of course you have those VMs configured with Fault Tolerance...
You might want to look at Proactive HA though: Configure Proactive HA
In terms of the cpu/memory question, you might want to look at Admission Control to ensure there is sufficient capacity in the event of a failure: vSphere HA Admission Control
So, If the host dies, VM dies, then VM migrate to another host, then reboots. The reboot requirement is to change the state/config of the VM (maybe on .vmx) from the dead host to another live host? Am I correct?
We're looking for a deeper explanation why the VM reboots on HA, we thought that there's a "vmotion" that can triggers the migration, why would vmotion do the trick.. I only know we can do vmotion when the hosts are up.
My apologies, cause we're having a hard time looking for answers like this in the internet. I appreciate for sharing your knowledge or experienced on this matter.
My colleague and I, will discuss to the client about Fault Tolerance since they have only 3 critical VMs.
Lets say a host "dies" by losing power, its no longer running so any VMs that were running on it have also stopped. Therefore the VM has to power on, on another host. You cant vMotion a VM if a host is offline or partitioned, just like you couldnt transfer a file to a server if it was powered off or partitioned from the network.
For an explination of HA, i highly recommend Clustering Deep Dive pages 53 onwards, it has a thorough explination of how and why HA works
You can FT Protect VMs which will protect you against losing a host without causing the VM to be powered off.
I get it now. I'm reading the ebook Deep Dive., and found what I'm looking for. Thank you for your time, I appreciate sharing your knowledge.