CONCURRENT ACCESS AND TRANSACTIONS IN A DISTRIBUTED FILE SYSTEM

• US 20160117336A1
• Filed: 10/27/2014
• Published: 04/28/2016
• Est. Priority Date: 10/27/2014
• Status: Active Grant

• Current Assignee: Cohesity, Inc.
• Original Assignee: Cohesity, Inc.
• Inventors: ARON, Mohit, SHANMUGANATHAN, Ganesha
• Granted Patent: US 9,697,227 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/10 File systems; File servers
• CPC Class Codes: G06F 16/1767 Concurrency control, e.g. optimistic or pessimistic approaches
• CPC Class Codes: G06F 16/182 Distributed file systems
• CPC Class Codes: G06F 16/1865 Transactional file systems

Abstract

Embodiments described herein provide techniques for maintaining consistency in a distributed system (e.g., a distributed secondary storage system). According to one embodiment of the present disclosure, a first set of file system objects included in performing the requested file system operation is identified in response to a request to perform a file system operation. An update intent corresponding to the requested file system operation is inserted into an inode associated with each identified file system object. Each file system object corresponding to the inode is modified as specified by the update intent in that inode. After modifying the file system object corresponding to the inode, the update intent is removed from that inode.

Claims

1. A method for performing a transaction in a distributed file system, the method comprising:

in response to a request to perform a file system operation, identifying a first set of file system objects modified in performing the requested file system operation;

inserting an update intent corresponding to the requested file system operation into an inode associated with each identified file system object, wherein the update intent specifies an ordered sequence of inodes associated with the requested file system operation; and

for each inode;

modifying either the inode or the file system object corresponding to the inode as specified by the update intent in that inode, and

after modifying the file system object corresponding to the inode, removing the update intent from that inode.

2. The method of claim 1, further comprising, prior to inserting the update intent corresponding to the requested file system operation into one of the inodes:

upon determining that a first one of the inodes stores an update intent from an incomplete file system operation, determining whether to complete or abandon the incomplete file system operation based on whether (i) a first set of one or more inodes at the beginning of the ordered sequence do not include an update intent related to the incomplete file system operation and (ii) a second set of one or more inodes following the first set each includes the update intent related to the incomplete file system operation.

3. The method of claim 2, wherein abandoning the incomplete file system operation comprises:

identifying, in the update intent stored in the first one of the inodes, the ordered sequence of inodes associated with the incomplete file system operation;

retrieving, as available, the inodes specified by the ordered sequence; and

upon determining that (i) the first set of one or more inodes at the beginning of the ordered sequence includes the update intent related to the incomplete file system operation and (ii) that the second set of one or more inodes following the first set do not include the update intent related to the incomplete file system operation, abandoning the incomplete file system operation by removing the update intent from each inode specified by the ordered sequence that includes the update intent.

4. The method of claim 2, wherein completing the incomplete file system operation comprises:

identifying, in the update intent stored in the first one of the inodes, the ordered sequence of inodes associated with the incomplete file system operation;

retrieving, as available, the inodes specified by the ordered sequence; and

upon determining that (i) the first set of one or more inodes at the beginning of the ordered sequence does not include the update intent related to the incomplete file system operation and that (ii) the second set of one or more inodes following the first set each includes the update intent related to the incomplete file system operation, completing the file system as specified by the update intent in the second set of nodes.

5. The method of claim 4, further comprising, removing the update intent from each of the second set of inodes.

6. The method of claim 1, wherein the update intent specifies an identifier corresponding to the requested file system operation, the operation to be performed, and an ordered sequence of the inodes modified in performing the requested operation.

7. The method of claim 1, further comprising, prior to inserting the update intent corresponding to the requested operation, obtaining a node-specific lock and a process-specific lock on each inode to be modified.

8. The method of claim 1, wherein the distributed file system is exposed to clients as an NFS, SMB, or CIFS mount point, and wherein the requested file system operation is an NFS, SMB, or CIFS operation from a client sent to a node of the distributed file system.

9. A non-transitory computer-readable medium storing instructions that, when executed by a processor, cause the processor to perform an operation for performing a transaction in a distributed file system, the operation comprising:

in response to a request to perform a file system operation, identifying a first set of file system objects modified in performing the requested file system operation;

inserting an update intent corresponding to the requested file system operation into an inode associated with each identified file system object, wherein the update intent specifies an ordered sequence of inodes associated with the requested file system operation; and

for each inode;

modifying either the inode or the file system object corresponding to the inode as specified by the update intent in that inode, and

after modifying the file system object corresponding to the inode, removing the update intent from that inode.

10. The computer-readable medium of claim 9, wherein the operation further comprises, prior to inserting the update intent corresponding to the requested file system operation into a first one of the inodes:

upon determining that a first one of the inodes stores an update intent from an incomplete file system operation, determining whether to complete or abandon the incomplete file system operation based on whether (i) a first set of one or more inodes at the beginning of the ordered sequence do not include an update intent related to the incomplete file system operation and (ii) a second set of one or more inodes following the first set each includes the update intent related to the incomplete file system operation.

11. The computer-readable medium of claim 9, wherein the update intent specifies an identifier corresponding to the requested file system operation, the operation to be performed, and the ordered sequence of the inodes modified in performing the requested operation.

12. The computer-readable medium of claim 9, wherein the operation further comprises, prior to inserting the update intent corresponding to the requested operation, obtaining a node-specific lock and a process-specific lock on each inode to be modified.

13. The computer-readable medium of claim 9, wherein the distributed file system is exposed to clients as an NFS, SMB, or CIFS mount point, and wherein the requested file system operation is an NFS, SMB, CIFS operation from a client sent to a node of the distributed file system.

14. A system comprising:

a processor; and

a memory storing program code, which, when executed on the processor performs an operation for performing a transaction in a distributed file system, the operation comprising;

in response to a request to perform a file system operation, identifying a first set of file system objects modified in performing the requested file system operation,

inserting an update intent corresponding to the requested file system operation into an inode associated with each identified file system object, wherein the update intent specifies an ordered sequence of inodes associated with the requested file system operation, and

for each inode;

modifying either the inode or the file system object corresponding to the inode as specified by the update intent in that inode, and

after modifying the file system object corresponding to the inode,

removing the update intent from that inode.

15. The system of claim 14, wherein the operation further comprises, prior to inserting the update intent corresponding to the requested file system operation into one of the inodes:

upon determining that a first one of the inodes stores an update intent from an incomplete file system operation, determining whether to complete or abandon the incomplete file system operation based on whether (i) a first set of one or more inodes at the beginning of the ordered sequence do not include an update intent related to the incomplete file system operation and (ii) a second set of one or more inodes following the first set each includes the update intent related to the incomplete file system operation.

16. The system of claim 15, wherein abandoning the incomplete file system operation comprises:

identifying, in the update intent stored in the first one of the inodes, the ordered sequence of inodes associated with the incomplete file system operation;

retrieving, as available, the inodes specified by the ordered sequence; and

upon determining that (i) the first set of one or more inodes at the beginning of the ordered sequence includes the update intent related to the incomplete file system operation and (ii) that the second set of one or more inodes following the first set do not include the update intent related to the incomplete file system operation, abandoning the incomplete file system operation by removing the update intent from each inode specified by the ordered sequence that includes the update intent.

17. The system of claim 15, wherein completing the incomplete file system operation comprises:

identifying, in the update intent stored in the first one of the inodes, the ordered sequence of inodes associated with the incomplete file system operation;

retrieving, as available, the inodes specified by the ordered sequence; and

upon determining that (i) the first set of one or more inodes at the beginning of the ordered sequence does not include the update intent related to the incomplete file system operation and that (ii) the second set of one or more inodes following the first set each includes the update intent related to the incomplete file system operation, completing the file system as specified by the update intent in the second set of nodes.

18. The system of claim 17, wherein the operation further comprises, removing the update intent from each of the second set of inodes.

19. The system of claim 14, wherein the operation further comprises, prior to inserting the update intent corresponding to the requested operation, obtaining a node-specific lock and a process-specific lock on each inode to be modified.

20. The system of claim 14, wherein the distributed file system is exposed to clients as an NFS, SMB, or CIFS mount point, and wherein the requested file system operation is an NFS, SMB, or CIFS operation from a client sent to a node of the distributed file system.

21. A method for locking entities to be modified in an operation in a distributed file system, the method comprising:

receiving a request to perform a file system operation, wherein the request specifies one or more file system objects to be modified in a distributed file system managed by a cluster of nodes;

requesting, from a distributed lock manager, a node-specific lock on each of the file system objects, wherein the node-specific lock prevents other nodes in the cluster from accessing the locked file system objects during the file system operation; and

upon receiving the node-specific locks on the file system objects, requesting, from a process-specific lock manager, a process-specific lock on each of the file system objects, wherein the process-specific lock prevents other local processes from accessing the file system objects.

22. The method of claim 21, further comprising, performing the file system operation.

23. The method of claim 22, further comprising, upon completion of the file system operation, releasing the node-level locks and the process-level locks.