摘要 |
In a replicated database environment, all nodes (30) monitor the well-being of each other by receipt of node status messages over a network (20). When a node is detected as no longer being alive, the other nodes deduce that such node has failed. When a node failure is detected, a nodestatus record having a nodedown value is written into a node log (80) with respect to the failing node. Similarly, when a node comes alive or a new node becomes known to the network, a node status record having a nodeup value is written into the node log. When a local node fails and then is restarted, the local node uses the contents of its node log in order to determine how to initialize the value of a replicated data object. By reading its node log, the local node can decide whether the data object can be initialized from its disk (32) or not.
|