0465 Duplicate Processing¶
Since metadata is being created by many tools, it is possible that the same resource is catalogued multiple times. These duplicate instances are then exposed when the tools are connected together with Egeria. The duplicate processing types in this model are used by the governance engines and stewards to link entities that are duplicates so that the open metadata ecosystem can ensure consumers of open metadata do not see the duplicates when they query the open metadata repositories.
When a governance engine or steward detects that two elements are duplicates of one another, they are linked using the PeerDuplicateLink. This link on its own indicates a potential duplicate.
When the duplicate is confirmed (either by a steward, or a governance engine) the KnownDuplicate classification is added to the entities that are linked with the PeerDuplicateLink. This classification is detected by metadata retrieval processes and triggers the following duplicate processing:
When an entity is retrieved either by guid or via a query, and it has the KnownDuplicate classification attached, the most recently updated of the peer-linked entities is returned along with a combination of the classifications from itself and each of the entities linked with the PeerDuplicateLink relationship. If there are conflicts in the classifications they are logged in the audit log and the newest classification takes precedence.
When the relationships linked to an entity with the KnownDuplicate classification is requested, the relationships attached to this entity, along with the relationships attached to all entities linked with the PeerDuplicateLink relationship. If there should only be one of a particular type of relationship then the conflicts are logged to the audit log and the newest relationship takes precedence.
There is no special duplicate processing when a relationship is retrieved independently of an entity.
ConsolidatedDuplicate and ConsolidatedDuplicateLink¶
It may be that this simple set of survivorship rules and consolidation process is insufficient (or too expensive from a performance perspective). It is possible for a steward/governance engine to construct and store a consolidated entity with its consolidated classification and relationships. Such an entity is decorated with the ConsolidatedDuplicate classification and is linked to each of the source entities using the ConsolidatedDuplicateLink relationship. Once the ConsolidatedDuplicateLink relationship is in place, the simple survivorship rules and peer consolidation process are ignored and the properties, classifications and relationships from the consolidated entity are used for a query to one of the linked entities with the KnownDuplicate classification. The steward/governance engine is responsible for the ongoing maintenance of this consolidated entity.
Support for duplicate processing¶
The Stewardship Action OMAS provides APIs for setting up peer duplicates and consolidated entities.
The Governance Action Services running in the Governance Engines can automate the detection of duplicates and the maintenance of consolidated entities. The governance engines are supported by the Governance Engine OMAS.
- KnownDuplicateLink is deprecated in favor of the two specialized relationships: PeerDuplicateLink and ConsolidatedDuplicateLink.
Raise an issue or comment below