Method for identifying logical data discrepancies between database replicas in a database cluster using enhanced transaction logging

Abstract

A method and system for monitoring and maintaining the consistency of replicated databases in a shared-nothing database cluster architecture is presented. In order to improve the ability of the system to maintain data consistency between the various database replicas in the cluster, an enhanced relational database management system is described that: (a) tags each data change record in the transaction log for a given managed database with a unique transaction identifier that is associated with the transaction request that initiated the data change; and, (b) tags each data change record in the transaction log for a given managed database with a client identifier that identifies the client that submitted the transaction request that initiated the data change. The enhanced relational database management system also includes an extended client interface that makes the unique transaction identifier for each transaction request available to the client application that submitted the transaction request.

1 . A system for monitoring and maintaining data consistency comprising:
 a plurality of database nodes each capable of operating independently, each containing at least one database and each providing a transaction identifier associated with each database transaction request applied;   a transaction log for recording each insertion, deletion or modification of data for each of the at least one databases in the database nodes and for recording the transaction identifiers for each insertion, deletion or modification of data recorded in the transaction log;   a controller node in communication with the database nodes that can receive a series of database transaction requests from a client and that sends the series of database transaction requests to the plurality of database nodes and monitors the transaction identifiers from each of the plurality of database nodes.   whereby the transaction logs of each of the database nodes which are determined by the controller node to have completed the same database transactions are compared using the transaction identifiers to detect any inconsistency.   
     
     
         2 . The system of  claim 1  wherein a batch of related database transaction requests receives a single transaction identifier from each of the plurality of database nodes. 
     
     
         3 . The system of  claim 1  where if the controller node detects in a database's transaction log a transaction identifier not associated with a database transaction request sent by the controller, the controller applies a database transaction request to the database node reversing the changes associated with the transaction identifier.