System and method for use with a data grid cluster to support death detection

摘要

A system and method is described for use with a data grid cluster to support death detection. A network ring is formed by connecting a plurality of process nodes in the data grid, wherein each node in the network ring watches another node. A death of a first process node in the network ring can be detected by a second process node, when the second process node notices that its connection to the first process node has closed. The first process node then informs other process cluster nodes in the network ring that the first node is dead. In accordance with an embodiment, machine level death detection can also be supported in the data grid cluster by using an Internet Protocol (IP) monitor.

主权项

1. A method for supporting fast process level death detection in a data grid cluster comprising a plurality of process nodes operating on a plurality of machines, each machine comprising a microprocessor, a first subset of the plurality of process nodes having a first role and a second subset of the plurality of process nodes having a second role, the method comprising:
maintaining a death detection socket open on each of the plurality of process nodes during operation of said each of the plurality of process nodes; connecting the plurality of process nodes in the data grid cluster to form a main network ring, wherein each node in the main network ring watches another node in the main network ring; connecting the first subset of the plurality of process nodes having the first role to form a first secondary ring wherein each of the first subset of the plurality of process nodes watches another of the first subset of the plurality of process nodes in the first secondary ring; connecting the second subset of the plurality of process nodes having the second role to form a second secondary ring wherein each of the second subset of the plurality of process nodes watches another of the second subset of the plurality of process nodes in the second secondary ring; detecting a death of a first process node of the plurality of process nodes by a second process node of the plurality of process nodes, when the second process node notices that said death detection socket on the first process node is closed; and informing other process nodes of the plurality of nodes from the second process node that the first process node is dead.