Dispersion dependency in oracle clusterware

摘要

A method and apparatus for resource dispersion in a clusterware system is provided. The resource profiles for resources within a cluster of nodes are stored. Each resource profile includes attributes for a resource. One or more of the attributes in one of the resource profiles includes a dispersion dependency attribute that specifies that the resource is preferably not co-located with another resource. Resources are placed within the cluster based, at least in part, on the dispersion preference.

主权项

1. A method comprising:
storing resource profiles for resources within a cluster comprising a plurality of nodes,
wherein each resource profile of said resource profiles includes one or more resource profile attributes for a respective resource,wherein a resource profile of said resource profiles includes a dispersion dependency attribute that contains a dispersion relation specifying one or more specified resources, andwherein said dispersion relation specifies a preference to not co-locate one or more resources corresponding to the resource profile with the one or more specified resources on a node of the cluster plurality of nodes; and accessing, by a computing device configured to manage the cluster; determining, by the computing device, to place one or more resources corresponding to the resource profile on one or more particular nodes of the plurality of nodes,
wherein said determining comprises:
determining whether the preference specified by said dispersion relation can be satisfied,when said dispersion relation is not able to be satisfied, determining to place the one or more resources corresponding to the resource profile without regard to the dispersion relation, andwhen said dispersion relation is able to be satisfied, determining to place the one or more resources corresponding to the resource profile based on the dispersion relation; and causing, by the computing device, the one or more resources to be provisioned on the one or more particular nodes of the cluster;
wherein the method is performed by one or more computing devices.