Summary

  • 7 minutes

In this module, you learned the following principles and procedures about Active Directory sites and replication:

  • Topology mirrors the network. Map every routed subnet to one site, keep two sites per link, and set costs so the least-cost path matches reality. Let the KCC and ISTG own connection objects unless you have a documented reason not to.
  • Intra-site favors freshness; inter-site favors efficiency. Change notification and uncompressed traffic within a site; scheduled, compressed traffic between sites—optionally with inter-site change notification on fast links.
  • Windows Server 2025 makes DNS-based discovery mandatory in practice. BlockNetBIOSDiscovery is enabled by default, removing the NetBIOS/mailslot fallback. Inventory legacy DC Locator dependencies and prove DNS-only discovery works before you rely on it.
  • Metadata is the source of truth. USNs order changes; the high-watermark vector tracks progress with each partner; the up-to-dateness vector dampens redundant replication across the topology. Together they let you state exactly what each DC has received.
  • Deleted objects are replicated changes. Tombstone lifetime, the AD Recycle Bin, and deleted-object metadata govern recovery and are how you detect and remove lingering objects—never reconnect a DC that's been offline past the tombstone lifetime.
  • Validate before you change. A clean repadmin /replsummary and dcdiag run is the gate before raising a functional level or enforcing DNS-only discovery.
  • Replication Priority Boost is a precision tool. Use setPriorityBoost (read back via the parameterized msDS-PriorityBoost RootDSE attribute) to favor a specific partner and NC—WAN promotions, functional primary/backup sites—after you've cleared real failures, not as a substitute for correct topology.

Further reading