PREVENTING DATA LOSS IN A STORAGE SYSTEM

• US 20090125669A1
• Filed: 11/13/2007
• Published: 05/14/2009
• Est. Priority Date: 11/13/2007
• Status: Active Grant

• Current Assignee: Network Appliance Incorporated (NetApp, Inc.)
• Original Assignee: Network Appliance Incorporated (NetApp, Inc.)
• Inventors: Goel, Atul, Gupta, Ratnesh, Leong, James
• Granted Patent: US 8,074,019 B2
• Time in Patent Office: Days
• US Class Current: 711/103
• CPC Class Codes: G06F 11/1435 using file system or storage system metadata

Abstract

Storage servers use a fast, non-volatile or persistent memory to store data until it can be written to slower mass storage devices such as disk drives. If the server crashes before a write can complete, the data remains safely stored in non-volatile memory. If the data cannot be committed to disk when the server reboots (e.g. because the destination mass storage device is unavailable), it is stored in a file. When the disk reappears, the data in the file may be used to restore a file or file system on the disk to a consistent state.

Claims

1. A method comprising:

identifying uncommitted data records in a persistent store of a data storage server during system initialization of the data storage server;

saving each identified uncommitted data record onto a corresponding aggregate if said corresponding aggregate is accessible; and

storing each of said identified uncommitted data records that corresponds to an inaccessible aggregate in a file.

2. The method of claim 1 wherein an accessible aggregate is a Redundant Array of Independent Disks (“

RAID array”

) that is presently capable of storing data, and

an inaccessible aggregate is a RAID array that is presently incapable of storing data.

3. The method of claim 1, further comprising:

detecting an accessible aggregate;

determining whether the accessible aggregate was last shut down cleanly;

if the accessible aggregate was last shut down uncleanly, identifying a file containing uncommitted data records for said accessible aggregate; and

replaying uncommitted data records from said file onto said accessible aggregate.

4. The method of claim 3, further comprising:

testing said file to determine whether the uncommitted data records in the file are stale before the replaying operation.

5. The method of claim 1 wherein said file includes:

an identifier of said inaccessible aggregate;

an identifier of said persistent store; and

a sequence number of said inaccessible aggregate.

6. The method of claim 1 wherein the data storage server is a first data storage server, the method further comprising:

transmitting said file to a second, different data storage server.

7. The method of claim 1, further comprising:

storing said file on a backup medium.

8. The method of claim 1 wherein said persistent store is one of a non-volatile random access memory (“

NVRAM”

), a battery-backed random access memory (“

RAM”

), or a Flash memory.

9. A system comprising:

a first storage server including a persistent memory;

a second storage server;

a data switch to couple a plurality of mass storage devices to the first data server and the second data server, wherein a first group of the mass storage devices are controlled by the first storage server and a second group of the mass storage devices are controlled by the second storage server;

aggregate recovery logic in the first storage server to collect data records pertaining to the first group of mass storage devices from the persistent memory; and

aggregate recovery logic in the second storage server to apply the collected data records to the first group of mass storage devices after the second storage server takes control of the first group of mass storage devices.

10. The system of claim 9 wherein the first group of mass storage devices comprise a Redundant Array of Independent Disks (“

RAID Array”

).

11. The system of claim 9 wherein the data switch is a Fibre Channel switch.

12. The system of claim 9 wherein the persistent memory is one of a Flash memory, a battery-backed random access memory or a nonvolatile random access memory.

13. A method comprising:

booting a storage server;

detecting Redundant Array of Independent Disks (“

RAID array”

) data that was cached in a persistent memory of the storage server before the booting operation, the RAID array data to be stored on a mass storage aggregate that was available for writing before the booting operation but is not available for writing after the booting operation; and

storing the RAID array data in a file.

14. The method of claim 13, further comprising:

detecting that the mass storage aggregate has become available for writing;

checking that the file contains valid data for the aggregate; and

writing data from the file to the aggregate.

15. The method of claim 13, further comprising:

mirroring the file to a second storage server.

16. The method of claim 13, further comprising:

receiving a request to write client data into a file stored on the mass storage aggregate;

saving the client data in the persistent memory; and

transmitting a confirmation to the client before writing the client data into the file, wherein

the receiving, saving and transmitting operations occur before the booting operation.

17. A method comprising:

during a system initialization of a data storage server,

identifying uncommitted data records representing client write requests stored in a short-term persistent store of the data storage server,

for each identified uncommitted data record, determining whether a corresponding aggregate is accessible, and

for each said identified uncommitted data record that corresponds to an inaccessible aggregate, storing said data record in a file and deleting the data record from the short-term persistent store; and

when a corresponding aggregate that was inaccessible during the identifying operation becomes accessible, replaying the data records from the file to commit said data records to said accessible aggregate.

18. The method of claim 17 wherein each uncommitted data record corresponds to a Redundant Array of Independent Disks (“

RAID array”

) parity stripe associated with a client write request.

19. The method of claim 17 wherein the data storage server is a first data storage server, the method further comprising:

making the file available at a second data storage server; and

transferring control of the inaccessible aggregate to the second data storage server, wherein

the replaying operation is performed by the second data storage server.

20. The method of claim 17 wherein the inaccessible aggregate is a Redundant Array of Independent Disks (“

RAID Array”

) including a plurality of mass storage devices, where a subset of the plurality of mass storage devices is malfunctioning so that existing data can be read from the RAID array but new data cannot be written to the RAID array.