Multi-replica asynchronous table replication

• US 10,095,764 B2
• Filed: 09/30/2015
• Issued: 10/09/2018
• Est. Priority Date: 06/19/2015
• Status: Active Grant

• Current Assignee: SAP SE
• Original Assignee: SAP SE
• Inventors: Park, Chang Gyoo, Lee, Juchang, Kim, Kyu Hwan, Na, Hyoung Jun, Lee, Hyejeong
• Primary Examiner(s): MENG, JAU SHYA
• Application Number:
  US14/871,675
• Publication Number:
  US 20160371357A1
• Time in Patent Office: 1,105 Days
• Field of Search: None
• CPC Class Codes: G06F 11/1451 by selection of backup contents
• CPC Class Codes: G06F 16/2343 Locking methods, e.g. distributed locking or locking implementation details
• CPC Class Codes: G06F 16/2358 Change logging, detection, and notification replication G06F16/27
• CPC Class Codes: G06F 16/2365 Ensuring data consistency and integrity
• CPC Class Codes: G06F 16/2379 Updates performed during online database operations; commit processing
• CPC Class Codes: G06F 16/2471 Distributed queries
• CPC Class Codes: G06F 16/273 Asynchronous replication or reconciliation
• CPC Class Codes: G06F 16/285 Clustering or classification
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

Technologies are described for performing replication within a database environment. Where a database transaction is replicated at multiple replica nodes, a replica node is selected as a coordinator replica node for the transaction. The other replica node or nodes are designated as follower replica nodes for the transaction. A follower replica node sends the coordinator replica node a precommit notification when the follower replica node has precommitted the transaction. The coordinator replica node sends the follower replica node a postcommit notification to commit the transaction when the transaction has been precommitted by all of the replica nodes to which the transaction is to be replicated.

Claims

1. A method, implemented at least in part by a coordinator replica node comprising a processing unit and memory, the coordinator replica node being in communication with a follower replica node and a source node, for facilitating asynchronous replication within a database environment using data manipulation language (DML) statements, the method comprising:

asynchronously receiving a first DML statement for a first transaction from the source node, wherein the source node commits the first transaction without waiting for the first DML statement to be replayed by the coordinator replica node or the follower replica node, the first DML statement being designated for replication at the coordinator replica node and the follower replica node;

asynchronously receiving a second DML statement for a second transaction from the source node, wherein the source node commits the second transaction without waiting for the second DML statement to be replayed by the coordinator replica node or the follower replica node, the second DML statement being designated for replication at the coordinator replica node and the follower replica node;

executing the first DML statement;

precommitting the first transaction at the coordinator replica node;

without waiting for the commit of the first transaction, executing the second DML statement;

receiving a precommit notification for the first transaction from the follower replica node;

committing the first transaction after the first transaction has been precommitted by the coordinator replica node and the coordinator replica node has received the precommit notification for the first transaction from the follower replica node; and

after the first transaction has been committed at the coordinator replica node, sending a postcommit notification for the first transaction to the follower replica node.

2. The method of claim 1, wherein receiving the precommit notification for the first transaction comprises dequeuing an entry that includes the precommit notification for the first transaction from a precommit notification queue, and wherein the sending the postcommit notification for the first transaction comprises enqueuing an entry that includes the postcommit notification for the first transaction to a postcommit notification queue.

3. The method of claim 1, wherein the precommit notification for the first transaction includes a timestamp set by the follower replica node, and wherein the postcommit notification for the first transaction includes a timestamp set by the coordinator replica node.

4. The method of claim 3, wherein the first transaction comprises one or more additional follower replica nodes and the timestamp set by the coordinator replica node is set to be larger than a timestamp set by the follower replica node and any timestamps set by the one more additional follower replica nodes for the first transaction.

5. The method of claim 1, wherein the first transaction remains in doubt, but accessible for purposes of replays, at the follower replica node until receipt of the postcommit notification for the first transaction by the follower replica node, wherein an in doubt transaction is not visible to external readers.

6. The method of claim 1, wherein the coordinator replica node is a first replica node and the follower replica node is a second replica node and, for an arbitrary transaction, the first and second replica nodes may act as a coordinator replica node or a follower replica node, further comprising receiving from the source node a notification designating the first replica node as the coordinator replica node for the first transaction.

7. A server comprising a processing unit and memory, wherein the server is adapted to operate a follower replica node of a database system, the follower replica node being in communication with a coordinator replica node, wherein the memory comprises computer executable instructions that, when executed on the processing unit, cause the server to perform operations for facilitating asynchronous replication within a database environment using data manipulation language (DML) statements, the operations comprising:

asynchronously receiving a first DML statement for a first transaction originating at a source node, wherein the source node commits the first transaction without waiting for the first DML statement to be replayed by the coordinator replica node or the following replica node, the first DML statement being designated for replication at the coordinator replica node and the follower replica node;

asynchronously receiving a second DML statement for a second transaction originating at the source node, wherein the source node commits the second transaction without waiting for the second DML statement to be replayed by the coordinator replica node or the follower replica node;

executing the first DML statement;

precommitting the first transaction at the follower replica node;

sending a precommit notification for the first transaction to the coordinator replica node;

without waiting for the commit of the first transaction, executing the second DML statement;

receiving a postcommit notification for the first transaction from the coordinator replica node; and

committing the first transaction at the follower replica node after the follower replica node has received the postcommit notification for the first transaction from the coordinator replica node.

8. The server of claim 7, wherein the follower replica node continues to process one or more DML statements for one or more other transactions after sending the precommit notification for the first transaction and before receiving the postcommit notification for the first transaction.

9. The server of claim 7, wherein sending the precommit notification for the first transaction comprises enqueuing an entry that includes the precommit notification for the first transaction to a precommit notification queue, and wherein receiving the postcommit notification for the first transaction comprises dequeuing an entry that includes the postcommit notification for the first transaction from a postcommit notification queue.

10. The server of claim 7, wherein the precommit notification for the first transaction includes a timestamp set by the follower replica node, and wherein the postcommit notification for the first transaction includes a timestamp set by the coordinator replica node.

11. The server of claim 10, wherein the follower replica node is a first follower replica node, one or more additional follower replica nodes are associated with the first transaction, and the timestamp set by the coordinator replica node is set to be larger than the timestamp set by the first follower replica node and any other timestamp set by the one or more additional follower replica nodes for the transaction.

12. The server of claim 7, wherein the first transaction remains in doubt, but accessible for purposes of replays, at the follower replica node until receipt of the postcommit notification for the first transaction by the follower replica node, wherein an in doubt transaction is not visible to external readers.

13. The server of claim 7, wherein the follower replica node comprises a post-commit notification queue, but the follower replica node does not enqueue to the post commit notification queue for the first transaction.

14. The server of claim 7, wherein the server is a first server and is in communication with a second server comprising a processing unit and memory, the second server being adapted to operate the source node, the source node implementing a sender side of a replication manager, and the first server implements a receiver side of the replication manager.

15. One or more non-transitory tangible computer-readable storage media storing computer-executable instructions for causing a server programmed thereby to perform a method for facilitating asynchronous replication within a database environment using data manipulation language (DML) statements and implementing a replication manager of a database system, the replication manager being in communication with first and second replica nodes, the method comprising:

designating the first replica node as a transaction coordinator node for a first transaction comprising a first DML statement, wherein the second replica node is a follower node for the first transaction;

asynchronously sending the first DML statement for the first transaction from a source node to the first replica node and the second replica node, wherein the source node may commit the first transaction without waiting for the first DML statement to be executed at the first and second replica nodes;

designating the second replica node as a transaction coordinator node for a second transaction comprising a second DML statement, wherein the first replica node is a follower node for the second transaction;

asynchronously sending the second DML statement for the second transaction from the source node to the first replica node and the second replica node, wherein the source node may commit the second transaction without waiting for the second DML statement to be executed at the first and second replica nodes;

routing a precommit notification for the first transaction from the second replica node to the first replica node, the precommit notification for the first transaction indicating that the first transaction has been precommitted by the second replica node;

routing a postcommit notification for the first transaction from the first replica node to the second replica node;

routing a precommit notification for the second transaction from the first replica node to the second replica node, the precommit notification for the second transaction indicating that the second transaction has been precommitted by the first replica node; and

routing a postcommit notification for the second transaction from the second replica node to the first replica node.

16. The one or more non-transitory tangible computer-readable storage media of claim 15, wherein at least a portion of the replication manager is implemented on at least one of the first replica node and the second replica node.

17. The one or more non-transitory tangible computer-readable storage media of claim 15, wherein at least a portion of the replication manager is implemented in the source node, and wherein designating the first replica node or the second replica node as the transaction coordinator is carried out at the source node.

18. The one or more non-transitory tangible computer-readable storage media of claim 15, wherein the precommit notification for the first transaction includes a first timestamp set by the second replica node, and wherein the postcommit notification for the first transaction includes a second timestamp set by the first replica node.

19. The one or more non-transitory tangible computer-readable storage media of claim 18, wherein the first timestamp set by the first replica node is set to be larger than the second timestamp set by the second replica node and set by any of one or more additional follower replica nodes for the first transaction.

20. The one or more non-transitory tangible computer-readable storage media of claim 15, wherein the first transaction remains in doubt, but accessible for purposes of replays, at the second replica node until receipt of the postcommit notification for the first transaction by the second replica node, wherein an in doubt transaction is not visible to external readers.