We are thinking of implementing Hyper V Clusteriazation on Windows server 2012 servers. We aim to use the Live migration feature to eliminate downtime when one of the servers fails. My question is: Is live migration hot, that is if the server currently hosting the Cluster Hyper V suddenly shuts down( power failure or something), will the clustered hyper V resume working immedialtely on the second server without interruption, or it will shut down and then start on the second server.
LiveMigration and virtual machine failover are two different things. LiveMigration is for planned migrations of a virtual machine from one Hyper-V host to another with no downtime of the virtual machine or it's services and applications.
Failover of a virtual machine occurs when the host it is running on fails and the cluster restarts the virtual machine on another cluster host, in which case there is downtime for the virtual machine and it's services and applications. When a cluster host fails the state of virtual machines running on that host is lost.
Live migration: When you initiate live migration, the cluster copies the memory being used by the virtual machine from the current node to another node, so that when the transition to the other node actually takes place, the memory and state information is already in place for the virtual machine. The transition is usually fast enough that a client using the virtual machine does not lose the network connection. If you are using Cluster Shared Volumes, live migration is almost instantaneous, because no transfer of disk ownership is needed. A live migration can be used for planned maintenance but not for an unplanned failover.
This is a wonderfull idea. Really. Can you tell me how you will migrate the state of the running virtual machine off the failed server?
Live Migration takes the current state (memory) of a running virtual machine and moves it over to another host so that there is no downtime / restart.
Now, asuming host A fails. Like all failing servers it does so nicely (flame, explision, admin in front flying through the air - we all know how servers fail from various movies).
How are you going to copy the state of the running virtual machines off the destroyed server?
And that is where your idea falls apart. There is this thing called "time". Without a time machine live migration can not be used in case of an ordinary server failure.
Failing machine is downtime. Point. Do not want that - make a cluster aware application and run it as a cluster on multiple machines.
This can not be achieved with Windows Hyper-V Clustering. I understand that your system might be of a continuous data processing type like Data Historians which have services which need to be synchronized and will be affected if Host A turns off during a power failure (due to normal things IT folks face like power supply failure, UPS system failure etc.. no need for the admin to be flying through the air, explosions etc..) In that case there are several solutions from 3rd party applications that support these types of failure scenarios.