System and method for restoring a single file from a snapshot

• US 7,051,050 B2
• Filed: 03/19/2002
• Issued: 05/23/2006
• Est. Priority Date: 03/19/2002
• Status: Active Grant

• Current Assignee: Network Appliance Incorporated (NetApp, Inc.)
• Original Assignee: Network Appliance Incorporated (NetApp, Inc.)
• Inventors: Kahn, Andy C., Edwards, John K., Chen, Raymond C., Patel, Kayuri
• Primary Examiner(s): Le, Debbie M.
• Application Number:
  US10/100,948
• Publication Number:
  US 20030182253A1
• Time in Patent Office: 1,526 Days
• Field of Search: 707200-205, 707100-101, 711/122, 711161-162
• US Class Current: 707/679
• CPC Class Codes: G06F 16/10 File systems; File servers

Abstract

The present invention provides a system and method for restoring a single file from a snapshot without the need to copy every individual block or inode from the snapshot. A file restore process duplicates the inode of a file within the active file system and performs a reconciliation process between the blocks of the twin inode and the snapshot inode. If the file does not exist within the active file system, a new buffer tree is created that points to the data blocks stored in the snapshot.

Claims

1. A method for restoring a single file to an active file system from a snapshot, the method comprising the steps of:

determining a type of file;

performing, in response to determining that the type of file is a file that is absent from the active file system, an empty file routine;

performing, in response to determining that the type of file is a file that exists in the active file system and is a large file, a restore routine including a reconcilation process, the reconciliation process including the steps of,

a. creating a twin inode, the twin inode including a copy of block pointers associated with the file in the active file system;

b. creating a new inode associated with the file to be restored;

c. performig a reconciliation of the block pointers included in the twin inode with a set of block pointers included in the snapshot; and

performing, in response to determining that the type of file is a file that exists in the active file system and is small file, a conventional restore operation.

2. The method of claim 1 wherein the step of performing the empty file routine further comprises the steps of:

creating a buffer tree, the buffer tree pointing to a set of data blocks stored in the snapshot; and

writing the buffer tree to the active file system.

3. The method of claim 1 further comprising the step of:

generating a zombie file from the twin inode.

4. The method of claim 1 wherein the step of performing a reconciliation further comprises the steps of:

selecting a block pointer from the twin inode;

comparing the selected block pointer with a block pointer associated with the snapshot;

moving, in response to the selected block pointer matching the block pointer associated with the snapshot, the selected block pointer from the twin inode to the active file system; and

copying, in response to the selected block pointer differing from the block pointer associated with the snapshot and in response to the block pointer not being allocated, the block pointer associated with the snapshot to the active file system.

5. The method of claim 4 wherein the steps of moving the selected block pointer and copying the block pointer are logged in a nonvolatile storage medium wherein the steps can be played back at a later time.

6. A computer storing files on an active file system, the computer comprising:

means for determining a type of file;

means for performing, in response to determining that the type of file is a file that is absent from the active file system, an empty file routine;

means for performing, in response to determining that the type of file is a file that exists in the active file system and is a large file, a restore routine including a reconciliation process, the reconciliation process having,

a. means for creating a twin inode, the twin inode including a copy of block pointers associated with the file in the active file system;

b. means for creating a new inode associated with the file to be restored;

c. means for performing a reconciliation of the block pointers included in the twin inode with a set of block pointers included in the snapshot; and

means for performing, in response to determining that the type of file is a file that exists in the active file system and is a small file, a conventional restore operation.

7. The computer of claim 6 further comprising means for logging a set of steps associated with the reconciliation of block pointers in a nonvolatile storage device, whereby the steps can be played back at a later time.

23. The computer of claim 6 wherein the means for performing the empty file routine further comprises:

means for creating a buffer tree, the buffer tree pointing to a set of data blocks stored in the snapshot; and

means for writing the buffer tree to the active file system.

24. The computer of claim 6 further comprising:

means for generating a zombie file from the twin inode.

25. The computer of claim 6 wherein the means for performing a reconciliation further comprises:

means for selecting a block pointer from the twin inode;

means for comparing the selected block pointer with a block pointer associated with the snapshot;

means for moving, in response to the selected block pointer matching the block pointer associated with the snapshot, the selected block pointer from the twin inode to the active file system; and

means for copying, in response to the selected block pointer differing from the block pointer associated with the snapshot and in response to the block pointer not being allocated, the block pointer associated with the snapshot to the active file system.

8. In a storage operating system, a reconcilation process for restoring a single file to an active file system from a snapshot, wherein the file exists in the active file system and wherein the file is a large file or a small file, the reconciliation process comprising instructions for performing the steps of:

creating a twin inode, the twin inode including a copy of all blocks associated with the file in the active file system;

creating a new inode for the file to be restored;

comparing a block pointer from the twin inode with a block pointer associated with the snapshot;

moving, in response to the twin inode block pointer matching the block pointer associated with the snapshot, the twin inode block pointer to file to be restored in the active file system;

copying, in response to the twin inode block pointer differing from the block pointer associated with the snapshot and the block pointer associated with the snapshot not being allocated in the active file system, the block pointer associated with the snapshot to the file to be restored in the active file system; and

copying, in response to the block pointer being allocated in the active file system, a data block from the snapshot to the active file system.

9. A method for restoring a single file to an active file system from a snapshot, the method comprising the steps of:

creating a tracking entry, the tracking entry storing state information related to progress of restoring the single file;

determining a type of file;

performing, in response to determining that the type of file is a file that is absent from the active file system, an empty file routine;

performing, in response to determining that the type of file is a file that exists in the active file system and is a large file, a restore routine including a reconciliation process, the reconciliation process including the steps of,

a. creating a twin inode, the twin inode including a copy of block pointers associated with the file in the active file system;

b. creating a new inode associated with the file to be restored;

c. performing a reconciliation of the block pointers included in the twin inode with a set of block pointers included in the snapshot; and

performing, in response to determining that the type of file is a file that exists in the active file system and is a small file, a conventional restore operation.

10. A computer storing files on an active file system, the computer comprising:

means for determining a type of file;

means for creating a tracking entry, the tracking entry storing state information related to progress of restoring a single file from a snapshot;

means for performing, in response to determining that the type of file is a file that is absent from the active file system, an empty file routine;

means for performing, in response to determining that the type of file is a file that exists in the active file system and is a large file, a restore routine including a reconciliation process, the reconciliation process having,

a. means for creating a twin inode, the twin inode including a copy of block pointers associated with the file in the active file system;

b. means for creating a new inode associated with the file to be restored;

c. means for performing a reconciliation of the block pointers included in the twin inode with a set of block pointers included in the snapshot;

means for performing, in response to determining that the type of file is a file that exists in the active file system and is a small file, a conventional restore operation; and

wherein the state information stored in the tracking entry enables the means for performing a restore routine including a reconciliation process and the means for performing an empty file routine to resume operation after an interruption.

11. A method for restoring a single file in a volume having a plurality of files, comprising:

(A) maintaining the single file in an active file, the active file maintained in an active file system;

(B) choosing a snapshot of the file, the snapshot created at a first time, the snapshot having a snapshot inode;

(C) issuing a restore command at a second time later than the first time;

(D) creating, in response to the restore command, a new file in the active file system, the new file having a new file inode;

(E) determining if the snapshot file exists in the active file system,

(E.1) if the snapshot file does not exist in the active file system, then copying the snapshot inode to the new file inode, thereby restoring the data blocks pointed to by the snapshot file into the active file system;

(E.2) if the snapshot file does exist in the active file system and is a large file, performing a reconciliation process that includes,

(F) creating a twin inode of the file in response to the restore command;

(G) comparing entries in the twin inode with entries in the snapshot inode,

(G.1) if a twin inode entry matches a snapshot inode entry then pointing the new file inode entry to the twin inode entry,

(G.2) if a twin inode entry does not match a snapshot inode entry then allocating space in the active file system for the data blocks pointed to in the snapshot and pointing the new file inode entry to the snapshot inode entry.

12. The method of claim 11, further comprising:

directing changes to the single file, the changes arriving after the restore command, to the active file.

13. The method of claim 11, further comprising:

deleting a twin inode entry in response to the pointing a new file inode entry to the twin inode entry, wherein after the restoration is complete the twin inode will have only entries which have changed between the first time when the snapshot was created and the second time when the restore command was issued.

14. The method of claim 11, further comprising:

allocating space in the active file system for the data blocks of the snapshot.

15. A method for restoring a single file in a volume having a plurality of files, comprising:

(A) maintaining the single file in an active file, the active file maintained in an active file system;

(B) choosing a snapshot of the file, the snapshot created at a first time, the snapshot having a snapshot inode;

(C) issuing a restore command at a second time later than the first time;

(D) creating, in response to the restore command, a new file in the active file system, the new file having a new file inode;

(E) determining if the snapshot file exists in the active file system, and if the snapshot file does exist in the active file system and is a large file, performing a reconciliation process that includes,

(F) creating a twin inode of the file in response to the restore command;

(G) comparing entries in the twin inode with entries in the snapshot inode,

(G.1) and if a twin inode entry matches a snapshot inode entry then pointing the new file inode entry to the twin inode entry,

(G.2) and if a twin inode entry does not match a snapshot inode entry then allocating space in the active file system for the data blocks pointed to in the snapshot and pointing the new file inode entry to the snapshot inode entry.

16. A filing system having a plurality of files operated to restore a single file of a volume, comprising:

(A) means for maintaining the single file in an active file, the active file maintamed in an active file system;

(B) means for choosing a snapshot of the file, the snapshot created at a first time, the snapshot having a snapshot inode;

(C) means for issuing a restore command at a second time later than the first time;

(D) means for creating, in response to the restore command, a new file in the active file system, the new file having a new file inode;

(E) means for determining if the snapshot file exists in the active file system,

(E.1) if the snapshot file does not exist in the active file system, then means for copying the snapshot inode to the new file inode, thereby restoring the data blocks pointed to by the snapshot file into the active file system;

(E.2) if the snapshot file does exist in the active file system and is a large file, performing a reconciliation process that includes,

(F) means for creating a twin inode of the file in response to the restore command;

(G) means for comparing entries in the twin inode with entries in the snapshot inode,

(G.1) if a twin inode entry matches a snapshot inode entry then means for pointing the new file inode entry to the twin inode entry,

(G.2) if a twin inode entry does not match a snapshot inode entry then means for allocating space in the active file system for the data blocks pointed to in the snapshot and means for pointing the new file inode entry to the snapshot inode entry.

17. The filing system of claim 16, further comprising:

means for directing changes to the single file, the changes arriving after the restore command, to the active file.

18. The filing system of claim 16, further comprising:

means for deleting a twin inode entry in response to the pointing a new file inode entry to the twin inode entry, wherein after the restoration is complete the twin inode will have only entries which have changed between the first time when the snapshot was created and the second time when the restore command was issued.

19. The filing system of claim 16, further comprising:

means for allocating space in the active file system for the data blocks of the snapshot.

20. A filing system having a plurality of files operated to restore a single file of a volume, comprising:

(A) means for maintaining the single file in an active file, the active file maintamed in an active file system;

(B) means for choosing a snapshot of the file, the snapshot created at a first time, the snapshot having a snapshot inode;

(C) means for issuing a restore command at a second time later than the first time;

(D) means for creating, in response to the restore command, a new file in the active file system, the new file having a new file inode;

(E) means for determining if the snapshot file exists in the active file system, and if the snapshot file does exist in the active file system and is a large file, performing a reconciliation process that includes,

(F) means for creating a twin inode of the file in response to the restore command;

(G) means for comparing entries in the twin inode with entries in the snapshot inode,

(G.1) and if a twin inode entry matches a snapshot inode entry then means for pointing the new file inode entry to the twin inode entry,

(G.2) and if a twin inode entry does not match a snapshot inode entry then means for allocating space in the active file system for the data blocks pointed to in the snapshot and means for pointing the new file inode entry to the snapshot inode entry.

21. A computer readable media, comprising:

said computer readable media containing instructions for execution on a processor for the practice of a method for restoring a single file in a volume having a plurality of files, having the steps of,

(A) maintaining the single file in an active file, the active file maintained in an active file system;

(B) choosing a snapshot of the file, the snapshot created at a first time, the snapshot having a snapshot inode;

(C) issuing a restore command at a second time later than the first time;

(D) creating, in response to the restore command, a new file in the active file system, the new file having a new file inode;

(E) determining if the snapshot file exists in the active file system,

(E.1) if the snapshot file does not exist in the active file system, then copying the snapshot inode to the new file inode, thereby restoring the data blocks pointed to by the snapshot file into the active file system;

(E.2) if the snapshot file does exist in the active file system and is a large file, performing a reconciliation process that includes,

(F) creating a twin inode of the file in response to the restore command;

(G) comparing entries in the twin inode with entries in the snapshot inode,

(G.1) if a twin inode entry matches a snapshot inode entry then pointing the new file inode entry to the twin inode entry,

(G.2) if a twin inode entry does not match a snapshot inode entry then allocating space in the active file system for the data blocks pointed to in the snapshot and pointing the new file inode entry to the snapshot inode entry.

22. A computer readable media, comprising:

said computer readable media containing instructions for execution on a processor for the practice of a method for restoring a single file in a volume having a plurality of files, having the steps of,

(A) maintaining the single file in an active file, the active file maintained in an active file system;

(B) choosing a snapshot of the file, the snapshot created at a first time, the snapshot having a snapshot inode;

(C) issuing a restore command at a second time later than the first time;

(D) creating, in response to the restore command, a new file in the active file system, the new file having a new file inode;

(E) determining if the snapshot file exists in the active file system, and if the snapshot file does exist in the active file system and is a large file, performing a reconciliation process that includes,

(F) creating a twin inode of the file in response to the restore command;

(G) comparing entries in the twin inode with entries in the snapshot inode,

(G.1) and if a twin inode entry matches a snapshot inode entry then pointing the new file inode entry to the twin inode entry,

(G.2) and if a twin inode entry does not match a snapshot inode entry then allocating space in the active file system for the data blocks pointed to in the snapshot and pointing the new file inode entry to the snapshot inode entry.