File system snapshots over fully provisioned volume file in direct mode

• US 9,256,614 B1
• Filed: 06/28/2013
• Issued: 02/09/2016
• Est. Priority Date: 06/28/2013
• Status: Active Grant

• Current Assignee: Emc IP Holding Company LLC (Dell Technologies Inc.)
• Original Assignee: EMC Corporation (Dell Technologies Inc.)
• Inventors: Bono, Jean-Pierre, Davenport, William C., de Forest, Miles A., Forrester, Walter C., Marko, Michal, Zhang, Ye, Armangau, Philippe
• Primary Examiner(s): Stevens, Robert
• Application Number:
  US13/931,764
• Time in Patent Office: 956 Days
• Field of Search: 707/827, 707/756
• US Class Current: 1/1
• CPC Class Codes: G06F 16/128 Details of file system snapshots on the file-level, e.g. snapshot creation, administration, deletion use of snapshots for error detection or correction G06F11/14, G06F11/16
• CPC Class Codes: G06F 16/22 Indexing; Data structures therefor; Storage structures
• CPC Class Codes: G06F 16/2228 Indexing structures

Abstract

A data storage system employs a multi-level stack including a lower-deck file system and an upper-deck file system providing a file system interface to a file system consumer. The lower-deck file system presents a volume file to provide storage to the upper-deck file system. Storage is reserved to the upper-deck file system, and the volume file is fully provisioned so an out-of-space condition is not presented to the upper-deck file system. In an initial direct operating mode a block address in the lower-deck file system is obtained by a calculation from a corresponding block address in the volume file. Snapshots of a primary volume file are created and stored in respective snapshot volume files. The lower-deck file system transitions into a mapped mode using generally arbitrary mappings of block addresses of the primary volume file and snapshot volume files to obtain corresponding block addresses in the lower-deck file system.

Claims

1. A data storage system for use by a file system consumer according to a file system interface, comprising:

one or more physical, non-transitory, nonvolatile storage devices; and

a processing subsystem executing software to form a multi-level stack of functional components including a lower-deck file system and an upper-deck file system, the upper-deck file system providing the file system interface to the file system consumer, the lower-deck file system presenting a volume file from which storage is provided to the upper-deck file system, units of storage of the storage devices being reserved to the upper-deck file system, the volume file being fully provisioned based on the reserved units of storage so that an out-of-space condition is not presented to the upper-deck file system, the reserved units of storage being added dynamically to the volume file to increase its allocated size in response to demand from the upper-deck file system, the lower-deck file system initially operating in a direct mode in which a block address of a data block in the lower-deck file system is obtained by a predetermined calculation from a corresponding block address of a data block in the volume file,

wherein the volume file is a primary volume file, and one or more snapshots of the primary volume file are created and stored in respective snapshot volume files of the lower-deck file system, each snapshot being a respective point-in-time copy of the primary volume file, and wherein the lower-deck file system transitions into a mapped mode of operation upon creation of the snapshots, the mapped mode including use of generally arbitrary mappings of respective block addresses of data blocks of the primary volume file and snapshot volume files to obtain corresponding block addresses of the data blocks in the lower-deck file system.

2. A data storage system according to claim 1, wherein upon creation of a snapshot it is populated with existing blocks of the primary volume file, and in response to ongoing writes to the primary file, a write split is performed by which current data blocks of the primary volume file are separated from the existing blocks which continue to populate the snapshot, the current data blocks being the subject of the ongoing writes as well as reads directed to the primary volume file.

3. A data storage system according to claim 2, wherein in response to deletion of a last one of the snapshots, the data blocks of the primary volume file are configured in a linear manner and the lower-deck file system transitions back into the direct mode with respect to the primary volume file.

4. A data storage system according to claim 1, wherein the snapshot volume files are first-level snapshot volume files, and wherein one or more second-level snapshots of the one of the first-level snapshot volume files are stored in respective second-level snapshot volume files of the lower-deck file system, each second-level snapshot being a respective point-in-time copy of a respective first-level snapshot volume file, and the mapped mode includes use of generally arbitrary additional mappings of respective block addresses of data blocks of the second-level snapshot volume files to obtain corresponding block addresses of the data blocks in the lower-deck file system.

5. A data storage system according to claim 1, wherein the mapped mode includes traversal of pointers and indirect blocks of an inode structure identifying locations of data blocks in the lower-deck file system based on block addresses of the data blocks in the volume file.

6. A data storage system according to claim 1, wherein:

all metadata of the upper-deck file system is stored within the volume file and therefore automatically included as part of the snapshots without need for separately making the metadata stable at a time a snapshot is obtained; and

at a later time of mounting a snap of the upper-deck file system, state is recovered by reading and applying the following from the metadata of the snap;

(1) upper deck slice pool state, (2) metadata state, and (3) file cache state.

7. A data storage system according to claim 1, wherein the units of storage are provided by a pool layer, and the multi-level stack of functional components includes a sparse volume layer of mapping between the lower-deck file system and the pool layer, the sparse volume layer including a sparse volume on which the lower-deck file system is structured and which is accessed by file operations of the lower-deck file system.

8. A data storage system according to claim 7, wherein the upper-deck file system draws slices of storage from the volume file, and the volume file is constituted by slices from the pool via operation of the lower-deck file system which draws slices from the pool.

9. A data storage system according to claim 1, wherein the upper-deck file system operates in a respective mapped mode including use of generally arbitrary mappings of respective block addresses of data blocks of files of the upper-deck file system to obtain corresponding block addresses of the data blocks of the volume file.

10. A method of operating a data storage system to provide file system functionality to a file system consumer according to a file system interface, the data storage system including one or more physical nonvolatile storage devices and a processing subsystem, comprising:

instantiating, in the processing subsystem, software to form a multi-level stack of functional components including a lower-deck file system and an upper-deck file system, the upper-deck file system providing the file system interface to the file system consumer; and

by the lower-deck file system, (i) presenting a volume file from which storage is provided to the upper-deck file system, units of storage of the storage devices being reserved to the upper-deck file system, the volume file being fully provisioned based on the reserved units of storage so that an out-of-space condition is not presented to the upper-deck file system, (ii) dynamically adding the reserved units of storage to the volume file to increase its allocated size in response to demand from the upper-deck file system, and (iii) utilizing a direct operating mode in which a block address of a data block in the lower-deck file system is obtained by a predetermined calculation from a corresponding block address of a data block in the volume file,

wherein the volume file is a primary volume file, and one or more snapshots of the primary volume file are created and stored in respective snapshot volume files of the lower-deck file system, each snapshot being a respective point-in-time copy of the primary volume file, and wherein the lower-deck file system transitions into a mapped mode of operation upon creation of the snapshots, the mapped mode including use of generally arbitrary mappings of respective block addresses of data blocks of the primary volume file and snapshot volume files to obtain corresponding block addresses of the data blocks in the lower-deck file system.

11. A method according to claim 10, wherein upon creation of a snapshot it is populated with existing blocks of the primary volume file, and in response to ongoing writes to the primary file, a write split is performed by which current data blocks of the primary volume file are separated from the existing blocks which continue to populate the snapshot, the current data blocks being the subject of the ongoing writes as well as reads directed to the primary volume file.

12. A method according to claim 11, wherein in response to deletion of a last one of the snapshots, the data blocks of the primary volume file are configured in a linear manner and the lower-deck file system transitions back into the direct mode with respect to the primary volume file.

13. A method according to claim 10, wherein the snapshot volume files are first-level snapshot volume files, and wherein one or more second-level snapshots of the one of the first-level snapshot volume files are stored in respective second-level snapshot volume files of the lower-deck file system, each second-level snapshot being a respective point-in-time copy of a respective first-level snapshot volume file, and the mapped mode includes use of generally arbitrary additional mappings of respective block addresses of data blocks of the second-level snapshot volume files to obtain corresponding block addresses of the data blocks in the lower-deck file system.

14. A method according to claim 10, wherein the mapped mode includes traversal of pointers and indirect blocks of an inode structure identifying locations of data blocks in the lower-deck file system based on block addresses of the data blocks in the volume file.

15. A method according to claim 10, wherein:

all metadata of the upper-deck file system is stored within the volume file and therefore automatically included as part of the snapshots without need for separately making the metadata stable at a time a snapshot is obtained; and

at a later time of mounting a snap of the upper-deck file system, state is recovered by reading and applying the following from the metadata of the snap;

(1) upper deck slice pool state, (2) metadata state, and (3) file cache state.