| 摘要 |
Apparatus and methods for detecting failure of a node in a cluster computer system. The apparatus include controllers programmed to cooperate. In one embodiment, the apparatus include first and second nodes with respective bus controllers communicatively coupled to each other and to a logical I/O device by means of a bus. The first node firstly recognizes the the second node as a node in the cluster. At some later point, a node whose communicative coupling has failed node is coupled to the cluster a second time. The first node recognizes this second coupling and, in response, queries the second node for failure-status information. The first and second nodes negotiate membership in the cluster for the second node on the first node's determining that the second node was the node that failed between the first and second couplings. Various embodiments follow: The first node firstly recognizes the second node as either a master or a slave node in the cluster. Then the negotiating of the second node's cluster membership includes ceasing to recognize the second node as the master node, negotiating slave membership in the cluster for the second node and thirdly recognizing the first node as the master node, on determining on the first node that the second node failed between the first and second couplings and that the second node was firstly recognized as the master node.
|
