Fix ERROR_DS_DRA_SOURCE_IS_PARTIAL_REPLICA (0x00002111)

Quick Answer

Run repadmin /syncall /AdeP from a writable domain controller, then use ntdsutil to remove the partial replica naming context if it persists.

Why This Happens

You're staring at ERROR_DS_DRA_SOURCE_IS_PARTIAL_REPLICA (0x00002111) in your replication logs. I've seen this most often when a domain controller that was demoted or is running as a read-only DC (RODC) gets mistakenly targeted as a replication source. The source DC only holds a partial replica of the directory partition — maybe it's an application partition that wasn't fully replicated, or the DC was originally a global catalog that lost that role.

This tripped me up the first time too. The error's exact wording: The replication source does not have a full replica of the directory partition. Translation: the DC you're pulling from doesn't own the full copy. It's like trying to borrow a book from someone who only has the dust jacket.

Fix Steps

Identify the culprit partition. Open Event Viewer on the destination DC, look for NTDS Replication events with ID 1925 or 1988. Note the partition DN (like DC=DomainDnsZones,DC=contoso,DC=com).
Check which DCs hold full replicas. On any DC, run:
```
repadmin /showrepl * /csv | findstr /i "partial"
```
This lists all partial replicas. Note the source DC that's failing.
Force sync from a known good writable DC. Run this on the destination DC:
```
repadmin /syncall /AdeP
```
Replace the default source with a specific DC if needed:
```
repadmin /sync    /force
```
If that fails, remove the bad replication link. Open ADSI Edit (or use ntdsutil). Connect to the destination DC's NTDS Settings. Under CN=NTDS Settings,CN=ServerName,CN=Servers,CN=SiteName,CN=Sites,CN=Configuration,..., find the CN=DomainDnsZones,CN=Partitions,... object. Right-click and delete the msDS-NC-Replica-Locations value that points to the partial source. Then force a fresh sync from a writable DC.
Clean up the partial replica reference. Use ntdsutil to remove the naming context from the broken DC:
```
ntdsutil
activate instance ntds
partition management
remove nc replica   /force
quit
```

Alternative Fixes

Scenario A: The source DC is an RODC. You can't replicate from an RODC to a writable DC — that's by design. Force all replication to go through a writable DC. Set the RODC's replication source to a writable DC using:

repadmin /options  +DISABLE_INBOUND_REPL
repadmin /options  -DISABLE_INBOUND_REPL

Then sync from the writable DC.

Scenario B: The partition is an application partition (like ForestDnsZones). It may not be present on all DCs. Check if the partition is supposed to be replicated to that DC. If not, remove the replica altogether:

dnscmd  /ZoneDelete  /DsDel

Be careful — this removes the DNS zone if it's a DNS partition.

Scenario C: Replication metadata is corrupt. Run a metadata cleanup for the source DC:

ntdsutil
metadata cleanup
connections
connect to server 
quit
select operation target
list domains
select domain 0
list sites
select site 0
list servers in site
select server 
quit
remove selected server
quit

Prevention Tip

Monitor your replication topology. I use a weekly repadmin /replsummary to catch these partial replica issues early. If you're demoting a DC, make sure it's fully synced outbound first. Don't let RODCs hold application partition replicas unless you really need them — they're partial by nature. Also check that all global catalog servers have full replicas of all naming contexts. A single missing partition can cascade into this error across multiple DCs.

One more thing: if you're using Windows Server 2008 R2 or older, partial replica errors can be more stubborn because the replication engine handles them differently. Upgrade to 2012 R2 or later if you can — the replication consistency checks are way better.