Partager via


3.1.1.1.15 Replication Latency and Tombstone Lifetime

Replication latency is the delay between the time of an originating update to an NC and the time when this update is reflected in all replicas of that NC. Some updates are superseded before reaching all replicas, but for the purposes of this simplified definition, consider an attribute update that is not followed by other updates to that attribute for a long time.

Administrators of Active Directory control replication latency by setting several variables, specified in section 6.1 and section 6.2. These variables ultimately control the schedules used for scheduled replication, and they control the use of event-driven replication. Replication latency is not fully predictable in a real system, because it depends upon the volume of read requests to DCs, the volume of originating update requests to DCs, and the availability of DCs and communications links.

If the typical replication latency is larger than the tombstone lifetime (the value of the tombstoneLifetime attribute of the Directory Services object specified in section 6.1.1.2.4.1.1, interpreted as a number of days), some tombstones or recycled-objects will be garbage collected before they have replicated to every NC replica. As a result, some objects will never be deleted in some replicas. To restore consistency of object existence, an administrator cleans up such lingering objects with utility programs.