Locality and time based dependency relationships in clusters
A method, apparatus, and system are directed toward configuring a dependency relationship between resources in a cluster. A dependency relationship between a dependent in a first resource group and a dependee in a second resource group is declared. The dependency relationship might include a locality based qualifier and/or a time based qualifier. The locality based qualifier includes a Local Node, Any Node, or From Resource Group Affinity relationship. The time based dependency qualifier includes a Strong dependency, Weak dependency, Online Restart dependency, or Offline Restart dependency. The declaration might be made using a graphical user interface, property list, configuration file, or the like. A candidate node on which to activate the first resource group is determined. The dependent is brought online on the candidate node based on whether an instance of the dependee is online on a node specified by the locality based qualifier.
1. A method for configuring a dependency relationship between resources in a cluster, comprising:
declaring at least one relationship for a dependent resource in a first resource group,
wherein the at least one relationship includes a Local Node dependency relationship between the dependent resource in the first resource group and a first dependee resource in a second resource group, and
wherein the at least one relationship further includes an Any Node dependency relationship between the dependent resource in the first resource group and a second dependee resource in the second resource group;
determining a candidate node on which to activate the first resource group;
satisfying the Local Node dependency relationship for an instance of the dependent resource, if an instance of the first dependee resource is online on the candidate node;
satisfying the Any Node dependency relationship for the instance of the dependent resource, if an instance of the second dependee resource is online on any node in the cluster;
performing a check, after declaring, to determine if every relationship between every resource in the first resource group and any other resource maintains a directed acyclic relationship; and
activating, based on the check succeeding, the instance of the dependent resource on the candidate node, if every relationship of the at least one relationship is satisfied.
2. The method of claim 1 , wherein declaring further comprises:
receiving the at least one relationship from a graphical user interface, a property list, a configuration file, a configuration repository, or a command line interface; and
encoding the at least one relationship on a processor readable medium, wherein the at least one relationship includes information about the dependent resource, at least one dependee resource, and a locality based relationship qualifier.
3. The method of claim 1 , further comprising:
declaring an identity for every resource in the first resource group, wherein an instance of every resource in the first resource group is configured to be online on a same node.
4. The method of claim 1 , further comprising:
failing to satisfy the Local Node dependency relationship for the first dependee resource, if the dependee resource is offline on the candidate node; and
failing to satisfy the Any Node dependency relationship for the second dependee resource, if the dependee resource is offline on every node in the cluster.
5. A non-transitory computer readable medium comprising instructions which, when executed by a processor, perform a method for configuring a dependency relationship between resources in a cluster, comprising:
declaring at least one relationship for a dependent resource in a first resource group,
wherein the at least one relationship includes a Local Node dependency relationship between the dependent resource in the first resource group and a first dependee resource in a second resource group, and
wherein the at least one relationship further includes an Any Node dependency relationship between the dependent resource in the first resource group and a second dependee resource in the second resource group;
determining a candidate node on which to activate the first resource group;
satisfying the Local Node dependency relationship for an instance of the dependent resource, if an instance of the first dependee resource is online on the candidate node;
satisfying the Any Node dependency relationship for the instance of the dependent resource, if an instance of the second dependee resource is online on any node in the cluster;
performing a check, after declaring, to determine if every relationship between every resource in the first resource group and any other resource maintains a directed acyclic relationship; and
activating, based on the check succeeding, the instance of the dependent resource on the candidate node, if every relationship of the at least one relationship is satisfied.
6. The non-transitory computer readable medium of claim 5 , wherein declaring further comprises:
receiving the at least one relationship from a graphical user interface, a property list, a configuration file, a configuration repository, or a command line interface; and
encoding the at least one relationship on a processor readable medium, wherein the at least one relationship includes information about the dependent resource, at least one dependee resource, and a locality based relationship qualifier.
7. The non-transitory computer readable medium of claim 5 , further comprising instructions for:
declaring an identity for every resource in the first resource group, wherein an instance of every resource in the first resource group is configured to be online on a same node.
8. The non-transitory computer readable medium of claim 5 , further comprising instructions for:
failing to satisfy the Local Node dependency relationship for the first dependee resource, if the dependee resource is offline on the candidate node; and
failing to satisfy the Any Node dependency relationship for the second dependee resource, if the dependee resource is offline on every node in the cluster.