Backing up Virtual Machines

Backing up the Virtual Hard Drive files

The safest way to backup a virtual machine hard drive file is to shut down, or suspend the virtual machine before performing the backup.  Since virtual machine HDD files are typically large files, this can require substantial down time and it is not practical for all situations.   However, it should be noted that with Rhinoback, only the changed parts of virtual machine hard drive (HDD) file are backed up after the initial backup.  These incremental delta backups substantially reduce the amount of time required to backup the HDD file.

Virtual machines can be backed up using Rhinoback while they are running as long as an Open File agent is in use.  If the virtual machine is hosted on Windows 2003 Server, Windows XP, Windows Vista, or Windows 2008 Server, then the Volume Shadow Copy Service (VSS) that is provide as part of the operating system is automatically invoked.  This means that virtual machines hosted on one of the supported windows operating systems can be backed up with Rhinoback while they are running.  If the virtual machine is hosted on Linux or Mac machine, then a third party open file agent may be needed.  If you are runing VMWare ESX, then you can use the snapshot module (VCBMonitor.exe), that is part of the ESX Server.  Be aware that any method of backing up a Virtual Machine Hard Drive file (HDD or vmdk) while it is running will result in a backup that is essentially the same state as if the machine had been abruptly powered off.  In most cases, this doesn't cause a problem, but some applications are not designed to handle this situation.

A third option, which is a safe compromise between the two options above, is to shut down or suspend the virtual machine just long enough to get a snap shot.  This can be accomplished with ESX by using the VCBMonitor utility after the machine is shut down and then restarting the virtual machine.  In a windows environment, if the virtual machine is stopped or suspended, a VSS snapshot will be made automatically by Rhinoback when the backup job starts.  Once the shadow copy is created the virtual machine can be started again.  On restore, the virtual machine will be in the state that it was in when it was stopped or suspended.  This technique results in minimal downtime and a safe and consistent backup.   If the VM is suspended or in a saved state, you should also backup the state files along with the virtual disk file.

Backing up from within the VM guest operating system

You can always install the Rhinoback software on the guest operating system of the Virtual Machine.  This will allow you to backup any files within the guest VM.  This is a safe and simple solution, but it may not provide the complete disaster recovery that virtual machines afford.  In case of a disaster, you would have to recreate your virtual machine environment and then restore the data that was backed up fromt he Guest VM.

Other Considerations

As with backing up any large files using Rhinoback, the delta backup technology allows you to backup only the changed parts of the files.  This makes the backup's much faster and the require much less bandwidth.  However, you should consider running a full backup at regular intervals so that you don't accumulate excessive numbers of delta files.  You can set Rhinoback to automatically backup the entire file once a certain number of incremental delta versions are created, or you can specify that once a certain percentage of the data has changed then the entire file should be backed up again.  Keep in mind that more incremental deltas increase the amount of time required to perform a restore operation.  The advantage of schedule full backups on an interval rather than allowing Rhinoback to do a full backup when too many incrementals are accumulated is that you can schedule the full backups to occur at your choice of time, such as on Saturday night.

Overall, Rhinoback is excellent at backing up Virtual Machines and storing the images offsite.  This is a great solution to ensure that your VM's can be recovered in the event of disaster.