System and method for implementing a dead man dependency technique for cluster resources
View Patent ↗Method and system for implementing a dead man dependency technique in a cluster comprising a plurality of interconnected nodes, one of the nodes comprising a cluster resource manager (CRM), and one of the nodes comprising a policy engine (PE), the PE maintaining at least one dependency associated with a first resource executing on one of the nodes. In an embodiment, the method comprises, responsive to a status change of an attribute associated with the first resource, the PE receiving from the CRM an indication of the status change and information regarding a configuration of the cluster; responsive to receipt of the status change indication and the configuration information, identifying by the PE a manner of terminating and migrating the first resource to another one of the nodes; and providing by the PE to the CRM an instruction sequence for causing the identified manner of terminating and migrating the first resource.
1. A method for implementing a dead man dependency technique in a cluster comprising a plurality of interconnected nodes, one of the nodes comprising a cluster resource manager, and one of the nodes comprising a policy engine, the policy engine maintaining at least one dependency associated with a first resource executing on one of the nodes, the method comprising:
responsive to a status change of an attribute associated with the first resource, the policy engine receiving from the cluster resource manager an indication of the status change;
the policy engine receiving from the cluster resource manager information regarding a configuration of the cluster;
responsive to receipt of the status change indication and the configuration information, identifying by the policy engine a manner of terminating the first resource and migrating the first resource to another one of the nodes; and
providing by the policy engine to the cluster resource manager an instruction sequence for causing the identified manner of terminating the first resource and migrating the first resource to another one of the nodes.
2. The method of claim 1 wherein the attribute includes ownership of a token.
3. The method of claim 2 wherein the status change includes loss of ownership of the token.
4. The method of claim 1 wherein the attribute includes a connection between the node on which the first resource is executing and another node in the cluster.
5. The method of claim 4 wherein the status change includes loss of the connection.
6. The method of claim 1 wherein the configuration information includes configuration of the cluster around the time of the status change of the attribute associated with the first resource.
7. The method of claim 1 ,
wherein the manner of migrating the first resource to another node includes migrating the first resource to a geographically dispersed node;
wherein the dependency includes dependence on a second resource; and
wherein the manner of terminating the first resource includes simultaneously terminating the first resource and the second resource.
8. The method of claim 1 , wherein the policy engine is distributed among more than one of the nodes in the cluster.
9. A system for implementing a dead man dependency technique in a cluster comprising a plurality of interconnected nodes, one of the nodes comprising a cluster resource manager, and one of the nodes comprising a policy engine, the policy engine maintaining at least one dependency with a first resource executing on one of the nodes, the method comprising:
responsive to a status change of an attribute associated with the first resource, means for receiving from the cluster resource manager an indication of the status change;
means for the policy engine receiving from the cluster resource manager information regarding a configuration of the cluster;
responsive to receipt of the status change indication and the configuration information, means for identifying by the policy engine a manner of terminating the first resource and migrating the first resource to another one of the nodes; and
means for providing by the policy engine to the cluster resource manager an instruction sequence for causing the identified manner of terminating the first resource and migrating the first resource to another one of the nodes.
10. The system of claim 9 wherein the attribute includes ownership of a token.
11. The system of claim 10 wherein the status change includes loss of ownership of the token.
12. The system of claim 9 wherein the attribute includes a connection between the node on which the first resource is executing and another node in the cluster.
13. The system of claim 12 wherein the status change includes loss of the connection.
14. The system of claim 9 ,
wherein the configuration information includes configuration of the cluster around the time of the status change of the attribute associated with the first resource;
wherein the manner of migrating the first resource to another node includes means for migrating the first resource to a geographically dispersed node;
wherein the dependency includes dependence on a second resource; and
wherein the manner of terminating the first resource includes simultaneously terminating the first resource and the second resource.
15. A computer program product comprising computer-readable medium having stored thereon instructions executable by a computer for implementing a dead man dependency technique in a cluster comprising a plurality of interconnected nodes, one of the nodes comprising a cluster resource manager, and one of the nodes comprising a policy engine, the policy engine maintaining at least one dependency associated with a first resource executing on one of the nodes, the instructions executable by the computer for causing the computer to:
responsive to a status change of an attribute associated with the first resource, receive by the policy engine from the cluster resource manager an indication of the status change;
receive by the policy engine from the cluster resource manager information regarding a configuration of the cluster;
responsive to receipt of the status change indication and the configuration information, identifying by the policy engine a manner of terminating the first resource and migrating the first resource to another one of the nodes; and
provide from the policy engine to the cluster resource manager an instruction sequence for causing the identified manner of terminating the first resource and migrating the first resource to another one of the nodes.
16. The computer program product of claim 15 the attribute includes ownership of a token.
17. The computer program product of claim 16 wherein the status change includes loss of ownership of the token.
18. The computer program product of claim 15 wherein the attribute includes a connection between the node on which the first resource is executing and another node in the cluster.
19. The computer program product of claim 18 wherein the status change includes loss of the connection.
20. The computer program product of claim 19 ,
wherein the configuration information includes configuration of the cluster around the time of the status change of the attribute associated with the first resource;
wherein the manner of migrating the first resource to another node includes migrating the first resource to a geographically dispersed node;
wherein the dependency includes dependence on a second resource; and
wherein the manner of terminating the first resource includes simultaneously terminating the first resource and the second resource.