The basic idea behind instant recovery is that there is no need to restore a failed virtual machine from backup if a full copy of the VM exists within the backup storage array. The virtual machine can simply be run from backup storage, bringing it immediately back online. This is a relatively simple concept, but there are a lot of moving pieces involved with virtual machine data recovery.
Three key actions
Each vendor takes its own approach to instant recovery, but there are three general steps:
- The backup software must connect the VM to the backup copy of the virtual hard disk. Some backup applications modify the virtual machine so it points to a different storage location; other backup applications create a brand new VM that is connected to a backup copy of the virtual hard disk.
- Before the VM can be started, the backup needs to be protected. If you boot the virtual machine from backup storage without first implementing some safeguards, the backup copy of the VM will be modified. Write operations can then potentially undermine the integrity of the backup. As such, the backup application must take steps to prevent the contents of the backup from being modified.
- Before a VM can utilize backup storage, the backup software must create a snapshot of the backup copy of the virtual machine. The snapshot process creates a differencing disk that can intercept write operations for virtual machine data recovery. When the VM is booted from the backup storage array, any write operations are directed to the differencing disk, thereby leaving the backup in a pristine state. Some backup applications are able to continue making backups of the VM, even while it is running from the backup storage array.
These actions will bring the virtual machine online.
Possible pain points
It would not be prudent to allow a VM to run from backup storage on an indefinite basis. Doing so would cause two problems with virtual machine data recovery:
- Running VMs from your backup storage impacts performance. There is a limit to the number of IOPS the backup storage array can provide. The backups consume some IOPS and each virtual machine that runs from backup storage consumes additional IOPS. The greater the number of VMs being run from backup storage, the greater the potential for virtual machine performance to be negatively impacted.
- It can potentially impact your ability to back up those VMs. Some backup applications are incapable of backing up virtual machines that are already running from the backup storage array. Depending on how your backup system works and which vendor's product you use, you may even sacrifice VM high availability for any virtual machine running from the backup.
It is important to move the VM to a production virtualization host when it becomes possible to do so. The process typically uses a replication engine to copy the unmodified backup copy of the virtual machine to the production virtualization host.
Once this process has been completed, the contents of the differencing disk are merged into the production copy of the VM and the production and backup copies of the virtual machine are brought in sync with one another.
At that point, the hypervisor is redirected to use the production copy of the VM rather than the backup copy. Once control has been returned to the production VM copy, the virtual machine snapshot is deleted from the backup storage array. This deletes the VM's differencing disk, which is no longer needed for virtual machine data recovery.
How instant is instant recovery?
Instant virtual machine recoveries versus virtual test labs
Video: Guidelines for VM data recovery