Single step merging of a virtual hard disk chain
Embodiments for a method of backing up virtual hard disks by: parsing a list of base parent and differencing disks to identify disk parameters and child disks of immediate parent disks; creating a differencing disk chain in reverse time order of modification of the base parent disk starting from the base parent disk to a latest child differencing disk; identifying changed sectors in each child disk starting from the latest child differencing disk; creating a list detailing a respective differencing disk, sector offsets and logical data sector index for each changed sector; iteratively adding updated sectors to the list for latest changed sectors that were not previously added to the list by a later child disk by reading the changed sectors from their respective child disk; and merging, in a single step, the changed sectors read from their respective child disks into the base parent disk.
1. A computer-implemented method of backing up and recovering virtual hard disks in a large-scale data storage system, comprising:
creating, for a parent disk and set of differencing disks, a checkpoint chain identifying checkpoints set during snapshot backup operations on the virtual hard disks;
identifying changed sectors in each differencing disk starting with the latest differencing disk and proceeding in order back to the parent disk through each earlier differencing disk in a disk parsing process to identify changed sectors in each differencing disk relative to the parent disk;
storing the identified changed sectors for each differencing disk in a first set of respective entries in a table;
identifying a corresponding location of the identified changed sectors in the parent disk;
comparing changed sectors in each differencing disk with changed sectors in all earlier differencing disks to find matching changed sectors;
storing changed sectors for each differencing disk not matching changed sectors in all earlier differencing disks in a second set of respective entries in the table;
creating a final list of sectors to be updated with an identification of their corresponding differencing disk by combining all changed sectors from the second set of respective entries, wherein the final list of sectors contains only a latest change of the identified changed sectors for all the differencing disks to prevent redundant updates of sectors changed in two or more differencing disks; and
copying, in a single merge step using a hash table and a mapping using the second set of respective entities, the identified changed sectors of the final list of sectors and read from their respective differencing disks to a parent disk image stored on the backup server computer.
2. The method of claim 1 further comprising updating a sector bitmap and block allocation table (BAT) with the copied sectors through the mapping, and wherein for later disks in the copying step, sectors already identified as valid are not updated to the sector bitmap.
3. The method of claim 1 further comprising displaying, through a graphical user interface (GUI) the parent disk and the set of differencing disks in a hierarchical format.
4. The method of claim 1 wherein an immediately earlier differencing disk represents a parent disk to a current differencing disk, and further wherein the final list of sectors and corresponding differencing disks comprises a sequential map of updated sectors from latest to earliest in time.
5. The method of claim 1 wherein the backup server computer is part of a virtualized network comprising a hypervisor server computer and a plurality of virtual machine storage targets storing the parent disk as the virtual hard disks.
6. The method of claim 5 wherein the virtual hard disks are formatted in a VHD or VHDX format.
7. The method of claim 6 wherein each checkpoint comprises a hypervisor checkpoint constituting a respective differencing virtual hard disk.
8. The method of claim 7 wherein, once a checkpoint is taken, the respective differencing disk temporarily stores changes comprising sector updates or modifications to a host virtual machine original disk and rendering the parent disk as a read-only disk.
9. The method of claim 2 wherein the BAT comprises a table of absolute sector offsets to a file backing up the parent disk, and is pointed to by a table offset field of a dynamic disk header, and wherein the sector bitmap indicates which sectors contain valid data through a first bit state and which sectors have not been modified through a second bit state.
10. The method of claim 1 wherein the snapshot backup operation comprises a point-in-time, read-only backup of one or more files or volumes on the virtual hard disks, and wherein the backup operation comprises a full synthetic backup.
11. A system comprising a hardware processor for backing up and restoring virtual machines in a large-scale data storage system, comprising:
an identifying component creating, for a parent disk and set of differencing disks, a checkpoint chain identifying checkpoints set during snapshot backup operations on the virtual hard disks, and identifying changed sectors in each differencing disk relative to the parent disk starting with the latest differencing disk and proceeding, in a bottom-up approach, in order back to the parent disk through each earlier differencing disk, and identifying a corresponding location of the identified changed sectors in the parent disk, and storing the identified changed sectors for each differencing disk in a first set of respective entries in a table, comparing changed sectors in each differencing disk with changed sectors in all earlier differencing disks to find matching changed sectors, and storing changed sectors for each differencing disk not matching changed sectors in all earlier differencing disks in a second set of respective entries in the table;
a hardware client-side merge component creating, during a backup operation, a final list of sectors to be updated with an identification of their corresponding differencing disk by combining all changed sectors from the second set of respective entries, wherein the final list of sectors contains only a latest change of the identified changed sectors for all the differencing disks to prevent redundant updates of sectors changed in two or more differencing disks, and copying, in a single-step merge function using a hash table and a mapping using the second set of respective entities, the identified changed sectors of the final list of sectors and read from their respective differencing disks to a parent disk image stored on the backup server computer; and
an updating component updating a sector bitmap and block allocation table (BAT) with the copied sectors.
12. The system of claim 11 wherein an immediately earlier differencing disk represents a parent disk to a current differencing disk, and wherein the final list of sectors and corresponding differencing disks comprises a sequential map of updated sectors from latest to earliest in time.
13. The system of claim 12 wherein the backup server computer is part of a virtualized network comprising a hypervisor server computer and a plurality of virtual machine storage targets storing the parent disk as the virtual hard disks, and further comprising a hypervisor server, and wherein the virtual hard disks are formatted in a VHD or VHDX format, and wherein each checkpoint constitutes a respective differencing virtual hard disk.