Managing Restart Of Non-Recoverable Changes In A Replication System
Publication Date: 2014-Jun-19
The IP.com Prior Art Database
A method and system is disclosed for managing restart of non-recoverable or automatically committed changes that are managed in a replication system based on transactional data.
Page 01 of 3
Managing Restart Of Non -
-Recoverable Changes In A Replication System
Recoverable Changes In A Replication System
Disclosed is a method and system for managing restart of non-recoverable or automatically committed changes that are managed in a replication system based on transactional data.
Restarting a replication system is required in case of a planned outage or a system failure. In the case of recoverable resources within a replication system, a transaction or database manager ensures that an update within a transaction is applied atomically. However, in the case of non-recoverable resources, the transaction or database manager performs auto-commit processing without undo-logging. Accordingly, the changes made to the non-recoverable resources are irreversible in case of failure.
In an embodiment of the present invention, the method and system tracks a potential leading edge time for non-recoverable updates. The potential leading edge time is the time of the last non-recoverable change in a group of non-recoverable changes. The potential leading edge time can be stored as additional information in an opaque restart token. During restart, the potential leading edge is used to apply non-recoverable changes with greater tolerance until the potential leading edge is passed.
The following provide additional details regarding the above described method and system:
URTime = Time of the group of non-recoverable changes (time of newest change in group)
Leading_edge = time of the newest Unit of Recovery or transaction (UOR) completely processed
Potential_leading_edge = time of the newest change across all non-recoverable UORs that apply has begun for
1. When accepting a new UOR, determine the apply mode:
if URTime <= Leading_edge at restart
use the more tolerant apply mode for the entire group else if URTime <= Potential_leading_edge at restart
use the more tolerant apply mode for the entire group else
Page 02 of 3
use customer specified apply mode
2. Before starting the UOR, update the potential leading edge time, if necessary: if URTime <= current Leading_edge in memory follow normal recoverable restart management since the UOR is covered by parallel apply rules
else if URTime <= current Potential_leading_edge in memory follow normal r...