Parallel truncation point management for log based replication

摘要

Disclosed herein are system, method, and computer program product embodiments for replicating data in a distributed database system. Data containing a replicated truncation point associated with a replicating system is received via a data path. It can then be determined that the truncation point represents the point at which all data in a transaction log has been replicated (e.g., successfully or safely) and the transaction log can then be truncated at the truncation point (i.e., the data up to the truncation point deflected). Data containing an additional replicated truncation point associated with an additional replicating system via an additional data path may be received. It can then be determined that the additional replicated truncation point represents the point at which all data in the transaction log has been replicated and the transaction log can be then truncated at the additional replicated truncation point.

主权项

1. A method, comprising:
sending, by a plurality of sender threads, a request for truncation points to a plurality of replicating systems, wherein each of the plurality of replicating systems is associated with a corresponding sender thread of the plurality of sender threads; receiving, by the plurality of sender threads and in response to the sent request, data containing a plurality of replicated truncation points associated with the plurality of replicating systems via a plurality of data paths; selecting a safe point from among the plurality of received replicated truncation points, the safe point representing a point at which all data in a transaction log has been replicated by the plurality of replicating systems; and truncating the transaction log at the determined safe point.