Background movement of data between nodes in a storage cluster

• US 8,010,485 B1
• Filed: 08/16/2010
• Issued: 08/30/2011
• Est. Priority Date: 10/20/2005
• Status: Active Grant

• Current Assignee: AmZetta Technologies, LLC (American Megatrends International, LLC)
• Original Assignee: American Megatrends Incorporated
• Inventors: Balakrishnan, Narayanan, Subramanian, Srikumar, Chatterjee, Paresh, Mahalingam, Anandh
• Primary Examiner(s): Ali; Mohammad
• Assistant Examiner(s): Hasan; Syed H
• Application Number:
  US12/857,050
• Time in Patent Office: 379 Days
• Field of Search: None
• US Class Current: 707/609
• CPC Class Codes: G06F 3/061 Improving I/O performance
• CPC Class Codes: G06F 3/0647 Migration mechanisms
• CPC Class Codes: G06F 3/067 Distributed or networked storage systems, e.g. storage area networks [SAN], network attached storage [NAS]

Abstract

A method, system, apparatus, and computer-readable medium are described for the background movement of data between nodes in a storage cluster. According to one method, exception tables are generated that include data identifying the areas on each node that need to be migrated to another node and the areas on each node that are to be migrated from another node. The exception tables may be generated in response to the failure of a node in the storage cluster or in other circumstances. A background resynchronization thread utilizes the contents of the exception tables to migrate data between the nodes of the cluster, thereby synchronizing the data stored on the nodes. Input/output operations directed to an area that is to be migrated from another node are redirected to the other node in accordance with a timeout period associated with the input/output operations.

Claims

1. A method for moving data between nodes of a storage cluster, the method comprising:

creating a positive exception table in a first node of the storage cluster in response to determining that one of the nodes of the storage cluster has failed, the positive exception table identifying areas of the first node that need to be migrated to a destination node in the storage cluster;

creating a negative exception table in the destination node that identifies areas of the destination node that need to be migrated from the first node to the destination node;

executing a background process for migrating data from the first node to the destination node in accordance with contents of the positive exception table, the background process;

copying the data to be migrated to a buffer;

writing the data from the buffer to an area on the destination node; and

following completion of the write, updating the positive exception table in the first node and the negative exception table in the destination node to indicate that the area of the destination node does not need to be migrated; and

during the execution of the background process;

receiving an input/output (I/O) operation at the first node, the I/O operation directed to an area of the destination node;

determining based upon the positive exception table whether the area of the destination node identified in the I/O operation needs to be migrated from the first node to the destination node; and

either forwarding the I/O operation to the destination node in response to determining that the area of the destination node identified in the I/O operation does not need to be migrated from the first node, or fielding the I/O operation at the first node in response to determining that the area of the destination node identified in the I/O operation needs to be migrated from the first node to the destination node.

2. The method of claim 1, further comprising migrating data from the destination node to the first node to perform bi-directional migration of data.

5. The method of claim 1, wherein the I/O operation is associated with a first timeout period set by an initiator of the I/O operation and a second timeout period set in accordance with a lock on the area of the first node.

6. The method of claim 5, wherein when the second timeout period expires, the I/O operation is fielded by the first node and a data migration associated with the area of the first node is canceled.

7. The method of claim 5, wherein the second timeout period is less than the first timeout period.

3. The method of clam 1, further comprising locking the area of the first node containing the data to be migrated.

4. The method of claim 3, wherein the locking operates to queue any input/output (I/O) operations received for the area of the first node to be migrated until the migration of the area is completed.

8. A method for moving data between nodes of a storage cluster, the method comprising:

creating a positive exception table in a first node of the storage cluster in response to determining that one of the nodes of the storage cluster has failed, the positive exception table identifying areas of the first node that need to be migrated to a destination node in the storage cluster;

creating a negative exception table in the destination node that identifies areas of the destination node that need to be migrated from the first node to the destination node;

executing a background process for migrating data from the first node to the destination node in accordance with contents of the positive exception table, the background process;

copying the data to be migrated to a buffer;

writing the data from the buffer to an area on the destination node; and

following completion of the write, updating the positive exception table in the first node and the negative exception table in the destination node to indicate that the area of the destination node does not need to be migrated; and

during the execution of the background process;

receiving an input/output (I/O) operation at the destination node, the I/O operation directed to an area of the destination node;

determining based upon the negative exception table whether the area of the destination node identified in the I/O operation needs to be migrated from the first node to the destination node; and

either forwarding the I/O operation to the first node in response to determining that the area of the destination node identified in the I/O operation needs to be migrated from the first node or fielding the I/O operation at the destination node in response to determining that the area of the destination node identified in the I/O operation does not need to be migrated from the first node.

9. The method of claim 8, further comprising migrating data from the destination node to the first node to perform bi-directional migration of data.

10. The method of claim 8, further comprising locking the area of the first node containing the data to be migrated.

11. The method of claim 10, wherein the locking operates to queue any input/output (I/O) operations received for the area of the first node to be migrated until the migration of the area is completed.

12. The method of claim 8, wherein the I/O operation is associated with a first timeout period set by an initiator of the I/O operation and a second timeout period set in accordance with a lock on the area of the first node.

13. The method of claim 12, wherein the second timeout period is less than the first timeout period.

14. A computer system comprising two or more storage server computers, each storage server computer operating as a storage node, and each storage node configured to generate an exception table in response to determining that one of the nodes of the storage cluster has failed, the exception table identifying storage areas for migration to a destination node in the computer system, and to execute a background process for migrating contents of the areas of the storage nodes to the destination node according to the contents of a source exception table,

wherein a background process is configured to:

migrate data from the source storage node to a destination storage node according to the contents of the source exception table;

copy the data to be migrated to a buffer in accordance with the source exception table;

write the data from the buffer to an area on the destination storage node; and

following completion of the write, update the source exception table and a destination exception table in the destination storage node to indicate that the area of the destination storage node does not need to be migrated, and

wherein each storage node is further configured to;

receive an input/output (I/O) operation directed to an area of the destination node during the execution of the background process;

determine based upon the contents of the source exception table whether the area of the destination node identified in the I/O operation needs to be migrated from the storage node to the destination node; and

either forward the I/O operation to the destination node in response to determining that the area of the destination node identified in the I/O operation does not need to be migrated from the storage node, or field the I/O operation at the storage node in response to determining that the area of the destination node identified in the I/O operation needs to be migrated from the storage node to the destination node.

15. The computer system of claim 14, wherein each storage node is further configured to generate the exception table in response to determining that another of the storage nodes has failed.

16. The computer system of claim 15, wherein the exception table is defined as a bitmap.

17. The computer system of claim 15, wherein the background process is executed in response to determining that a failed storage node has been replaced and wherein the destination node comprises a replacement storage node for the failed storage node.