Method of managing nodes in computer cluster
View Patent ↗A method is described of managing nodes in a computer cluster comprising: each node repeatedly broadcasting a cluster summary message; a cluster coordinator node identifying failed nodes by analysing cluster summary messages received from other nodes in the cluster; and the cluster coordinator node broadcasting an updated cluster organization status, if failed nodes are identified. In at least preferred embodiments, the broadcasts can be transmitted using an ad-hoc wireless network.
1. A method of managing nodes in a computer cluster comprising:
each node repeatedly broadcasting a heartbeat message;
each node repeatedly receiving a plurality of the heartbeat messages;
each node repeatedly broadcasting a cluster summary message, wherein the cluster summary message enumerates the nodes from which a heartbeat message was received;
a cluster coordinator node identifying failed nodes by analyzing cluster summary messages received from other nodes in the cluster; and
the cluster coordinator node broadcasting an updated cluster organization status, if failed nodes are identified;
wherein the cluster coordinator node detects a failed node if the cluster coordinator node does not receive a heartbeat message from the failed node; and
wherein none of the cluster summary messages received by the cluster coordinator node includes the failed node.
2. A method as claimed in claim 1 wherein each nodes generates a cluster summary messages using cluster summary messages received from other nodes.
3. A method as claimed in claim 1 wherein the broadcasts are transmitted using an ad-hoc wireless network.
4. A method as claimed in claim 1 wherein each node broadcasts heartbeat messages comprising that node's identity and health status information.
5. A method as claimed in claim 1 wherein the health status information comprises memory status, central processing unit utilization and network interface card status.
6. A method as claimed in claim 1 wherein if the cluster coordinator node fails a next highest ranked node takes over as cluster coordinator.
7. A method as claimed in claim 1 comprising monitoring a transceiver within each node and, if the transceiver is detected to be faulty, closing a cluster service.
8. A computer cluster comprising:
a plurality of nodes, each node being arranged to repeatedly broadcast a heartbeat message, receive a plurality of heartbeat messages, broadcast a cluster summary message, wherein the cluster summary message enumerates the nodes from which a heartbeat message was received;
a cluster coordinator node for identifying failed nodes by analyzing cluster summary messages received from other nodes in the cluster, the cluster coordinator node being arranged to broadcast an updated cluster organization status, if failed nodes are identified;
wherein the cluster coordinator node detects a failed node if the cluster coordinator node does not receive a heartbeat message from the failed node; and
wherein none of the cluster summary messages received by the cluster coordinator node includes the failed node.
9. A computer cluster as claimed in claim 8 wherein each node is arranged to generate the cluster summary messages using cluster summary messages received from other nodes.
10. A computer cluster as claimed in claim 8 wherein each node comprises a transceiver for transmitting the broadcasts using an ad-hoc wireless network.
11. A computer cluster as claimed in claim 8 wherein each node broadcasts heartbeat messages comprising that node's identity and health status information.
12. A computer cluster as claimed in claim 11 wherein the health status information comprises memory status, central processing unit utilization and network interface card status.
13. A computer cluster as claimed in claim 8 wherein if the cluster coordinator node fails a next highest ranked node takes over as cluster coordinator.
14. A computer cluster as claimed in claim 8 comprising a transceiver device fault detection service for monitoring a transceiver within each node and, if the transceiver is detected to be faulty, closing a cluster service on the node.
15. A node for a computer cluster comprising:
means for each node repeatedly broadcasting a heartbeat message;
means for each node repeatedly receiving a plurality of the heartbeat messages;
means for repeatedly broadcasting a cluster summary message, wherein the cluster summary message enumerates the nodes from which a heartbeat message was received;
means for identifying failed nodes by analyzing cluster summary messages received from other nodes in the cluster, and for broadcasting an updated cluster organization status, if failed nodes are identified, when the node is acting as a cluster coordinator node
wherein the cluster coordinator node detects a failed node if the cluster coordinator node does not receive a heartbeat message from the failed node; and
wherein none of the cluster summary messages received by the cluster coordinator node includes the failed node.