| 摘要 |
A computing system in the form of a cluster of a number of multiprocessing nodes maintains, in a fault tolerant manner, a distribution of configuration data for each of the nodes so that each node has a database containing the configuration data associated with that node. The database, and therefore, the configuration data it contains, associated with any one node is substantially identical to that of any other node. A process running on one of the nodes is responsible for receiving a requests that require modification of the configuration data. Effecting changes to the configuration data, and therefore the distributed databases, includes the steps of first writing the requested change to a master audit log, distributing the change request to all nodes, receiving back from the nodes acknowledgement of the change request being effected at the acknowledging node, and then writing again to the master audit log that the change has been effected throughout the system. The master audit log thereby contains a reliable copy of the configuration data maintained in the database associated with each node of the cluster so that in the event any of the configuration data becomes corrupted, it can be replaced with correct data from the master audit log.
|
