Fix it now
Error 8418 is published as the replication operation failing because of a schema mismatch between the servers involved. Microsoft divides these into transient ones, which clear after a replication cycle following a schema update, and persistent ones, which mean something else is blocking replication and has to be found first.
netdom query fsmo
dsquery * cn=schema,cn=configuration,dc=example,dc=com -scope base -attr objectVersion
reg query "HKLM\System\CurrentControlSet\Services\NTDS\Parameters" /v SystemSchemaVersion
repadmin /showrepl /errorsonly
- Compare objectVersion across the forest. Every DC should return the same number, and Microsoft publishes the expected value per release: 87 for Windows Server 2016, 88 for 2019 and 2022, and 91 for 2025.
- Compare each DC’s objectVersion against its own
SystemSchemaVersionregistry value. The schema version is stored in both places and they should agree. - If a partner simply lags, pull the partition across:
repadmin /replicate <target> <schema master> "cn=schema,cn=configuration,dc=example,dc=com", then recheck. - If the source DC’s objectVersion and SystemSchemaVersion are both correct and destinations still report 8418, restart the source DC. Microsoft documents that a DC can fail to reload the in-memory schema after a schema update, and a restart is the fix.
- If the schema master itself reports a version lower than the release you are introducing, forest preparation never completed. Run
adprep /forestprepfrom the media of that version.
adprep /forestprep needs an account in Enterprise Admins, Schema Admins and Domain Admins of the domain that hosts the schema master. adprep /domainprep needs Domain Admins in the domain being prepared.
If the versions agree everywhere and replication runs, stop here. If 8418 persists with matching versions, the next section covers what else produces it.
Why it happens
The schema is a single forest-wide partition holding the definition of every class and attribute Active Directory knows about. Each DC keeps a full replica, but only the schema master holds a writable copy. When a source DC ships an update containing an attribute the destination has never heard of, there is nowhere sensible to put it, and failing the operation is safer than discarding data.
Microsoft splits 8418 into two categories, and the split matters more than anything else in this article. Transient mismatches appear and disappear across the forest following the replication topology, and resolve after one replication cycle per partner. They are normal after a schema update. Persistent ones do not resolve, and they mean there is an underlying issue preventing normal replication – lingering objects, database corruption, replication quarantine, DNS or RPC communication failures, or strict replication consistency enforcement. Treating a persistent 8418 as a convergence problem and forcing replication at it repeatedly will never work, because the thing blocking replication is not the schema.
The schema partition has simply not converged yet
You have this one if The schema master holds the newer version, one or more partners hold an older one, and this started shortly after a schema update.
- Read objectVersion on each DC:
dsquery * cn=schema,cn=configuration,dc=example,dc=com -scope base -attr objectVersion - Force the partition across:
repadmin /replicate <target> <schema master> "cn=schema,cn=configuration,dc=example,dc=com" - Check for a backlog on both ends and confirm the link works at all.
- Recheck with
repadmin /showrepl /errorsonly. This is the case that clears on its own; you are just hurrying it.
Within a site this converges in about a minute. Between sites it follows the site link schedule, so a default configuration can take hours. Compare objectVersion rather than guessing.
The versions agree and 8418 persists
You have this one if objectVersion and SystemSchemaVersion are correct on the source, and the destinations that use it as an inbound partner keep failing.
- Confirm both readings on the source: the objectVersion attribute on the schema container, and the
SystemSchemaVersionregistry value. - If both are correct, restart the source DC. A DC can fail to reload the in-memory schema after receiving a schema update, and a restart is the documented remedy.
- Do not keep forcing replication at it. That cannot clear an in-memory cache.
- After the restart, recheck the destinations before doing anything else.
This is the case that wastes the most time, because everything you can read says the schema is right – and it is. What is wrong is what the running process is holding.
Forest preparation never completed
You have this one if The schema master reports a lower objectVersion than the release you are promoting expects, and the adprep logs record a failure.
- Log on to the schema master with an account in Enterprise Admins, Schema Admins and Domain Admins of the domain that hosts the schema master. All three are required.
- Run
adprep /forestprepfrom the adprep folder on the media of the version you are introducing. - Run
adprep /domainprepin each domain, as a member of Domain Admins in that domain. - Re-read objectVersion on the schema master and let it replicate before promoting anything.
Current releases run preparation automatically during promotion when the credentials allow it. Run it by hand when the promoting account is deliberately less privileged, or when you want the schema change in its own change window.
Something else is blocking replication entirely
You have this one if 8418 persists, versions agree, restarting the source changed nothing, and other replication errors are present too.
- Stop treating it as a schema problem. Microsoft lists lingering objects, database corruption, replication quarantine, and DNS or RPC failures among the underlying issues that produce a persistent 8418.
- Check for lingering objects and quarantine first, because those have their own errors and their own remedies.
- Check DNS and RPC between the pair: name resolution for the partner’s GUID alias, and the endpoint mapper.
- If none of that explains it, collect the diagnostic output and open a case rather than continuing to force schema replication.
An object’s security descriptor is too large
You have this one if Event 1450 with error 1340 alongside the 8418, and the failure attaches to particular objects rather than to the partition.
- Read the published cause: the object’s nTSecurityDescriptor attribute is too large. Windows limits the attribute to 65,535 bytes.
- The descriptor holds access control entries set directly on the object and entries inherited from its parent, so both contribute.
- Reduce the number of ACEs on the object, working out which are direct and which are inherited and from where.
- If several objects are affected, script the size calculation rather than opening each one by hand.
This is the case where forcing schema replication can never succeed, because the schema was never the problem. It is worth recognising early.
Full reference
The schema version by release, as published
| Windows Server release | objectVersion |
|---|---|
| Windows 2000 | 13 |
| Windows Server 2003 | 30 |
| Windows Server 2003 R2 | 31 |
| Windows Server 2008 | 43 |
| Windows Server 2008 R2 | 47 |
| Windows Server 2012 | 56 |
| Windows Server 2012 R2 | 69 |
| Windows Server 2016 | 87 |
| Windows Server 2019 | 88 |
| Windows Server 2022 | 88 |
| Windows Server 2025 | 91 |
Note that 2019 and 2022 share 88. That is not an error in the table: introducing a 2022 DC into a 2019 forest does not raise the schema version, which surprises people who are expecting a number to move as evidence that preparation ran.
Reading the version, both places it lives
| Where | How to read it |
|---|---|
| Active Directory | dsquery * cn=schema,cn=configuration,dc=<domain>,dc=<tld> -scope base -attr objectVersion |
| Active Directory, as a file | ldifde -f schemaver.ldf -d cn=schema,cn=configuration,dc=<domain>,dc=<tld> -l objectVersion |
| Active Directory, by hand | ADSI Edit: the ObjectVersion attribute on the schema container |
| The registry | HKLM\System\CurrentControlSet\Services\NTDS\Parameters\SystemSchemaVersion |
Check both on the source DC when a mismatch persists. If the two disagree, or if both are correct while destinations still fail, you are in the in-memory schema case and the answer is a restart of the source rather than more replication.
adprep credentials, in full
| Command | Membership required |
|---|---|
adprep /forestprep |
Enterprise Admins, Schema Admins, and Domain Admins of the domain that hosts the schema master |
adprep /domainprep |
Domain Admins in the domain being prepared |
All three groups are needed for /forestprep, not two. An account holding only Schema Admins and Enterprise Admins gets a prerequisite verification failure, which reads like a broken command rather than a missing membership and sends people looking for the wrong thing.
Schema changes are additive and one way. There is no supported way to remove a class or attribute, or to roll the version back. Take a system state backup of the schema master before running adprep or any vendor schema extension, and confirm you can restore it.
When 8418 is not really about the schema
This is the part most treatments of the error miss, and it is the reason the article is written the way it is. Microsoft categorises persistent 8418s as pointing at an underlying issue that prevents normal replication, and lists them:
- Lingering objects, which block inbound replication of a whole partition under strict replication consistency.
- Database corruption, which Microsoft escalates rather than documenting a self-service remedy for.
- Replication quarantine, which is a different error with a different recovery.
- DNS or RPC communication failures between the pair, which have their own well-documented diagnostics.
- Strict replication consistency enforcement, which is the mechanism by which the first of these becomes an outage.
- An oversized nTSecurityDescriptor on a specific object, which surfaces as event 1450 with error 1340 and is fixed by reducing access control entries, not by touching the schema.
If you have a persistent 8418, work down that list before you force replication again. Forcing schema replication at a partition that is blocked by lingering objects will fail every time, and it will look like a schema problem every time.
The neighbouring codes
| Code | Symbolic name | Published text |
|---|---|---|
8418 |
ERROR_DS_DRA_SCHEMA_MISMATCH | The replication operation failed because of a schema mismatch between the servers involved |
8543 |
ERROR_DS_DRA_SCHEMA_CONFLICT | The replication operation could not be completed due to a schema incompatibility |
8544 |
ERROR_DS_DRA_EARLIER_SCHEMA_CONFLICT | The replication operation could not be completed due to a previous schema incompatibility |
8542 |
ERROR_DS_DRA_SCHEMA_INFO_SHIP | Schema information could not be included in the replication request |
Vendor schema extensions
Applications extend the schema too, and a half-completed extension produces the same family of errors as a half-completed forest preparation. Run the vendor’s schema step on the schema master itself where the product allows it, from an account in Schema Admins, and confirm both the forest objectVersion and the product’s own version marker read the same on every DC before continuing the installation. Running it elsewhere adds a replication hop before the change is even visible, which is how these end up half-done.
When a licence is the actual fix
In most cases this costs nothing, because the fix is replication and preparation you already own. It turns into a purchase in one situation: the DC that will not converge runs a Windows Server release that cannot hold the current schema, and no command changes that. It has to be replaced with a supported build, which needs a Windows Server licence. Arco supplies Windows Server 2025 Standard, and will check whether Software Assurance on your existing agreement already entitles you to the newer version before you spend anything. If the replacement is going onto a host that already runs other Windows virtual machines, ask about the edition too: Standard limits how many Windows virtual instances the licensed hardware may run and Datacenter does not, so the right answer depends on your density rather than on a rule of thumb.
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
8418 |
ERROR_DS_DRA_SCHEMA_MISMATCH: the replication operation failed because of a schema mismatch between the servers involved | Microsoft Learn |
8543 |
ERROR_DS_DRA_SCHEMA_CONFLICT: the replication operation could not be completed due to a schema incompatibility | Microsoft Learn |
8544 |
ERROR_DS_DRA_EARLIER_SCHEMA_CONFLICT: the replication operation could not be completed due to a previous schema incompatibility | Microsoft Learn |
8542 |
ERROR_DS_DRA_SCHEMA_INFO_SHIP: schema information could not be included in the replication request | Microsoft Learn |
Confirm the fix worked
- objectVersion on the schema container reads the same on every DC in the forest, and matches the published value for the release you are running.
SystemSchemaVersionin the registry agrees with objectVersion on each DC.repadmin /showrepl /errorsonlyreturns nothing for the schema partition.- An attribute added by the schema change is visible on the previously lagging DC.
- No new 8418 appears over the following day, so you know you fixed it rather than pushed one cycle through.
Questions people ask about this
Versions match everywhere and 8418 keeps coming back. What now?
If the source DC holds the correct version in both Active Directory and the registry, restart the source DC. Microsoft documents that a DC can fail to correctly reload the in-memory schema after receiving a schema update, and a restart is the fix. Forcing replication cannot clear a stale in-memory cache.
Which groups do I need to run adprep?
For /forestprep: Enterprise Admins, Schema Admins, and Domain Admins of the domain that hosts the schema master. All three. For /domainprep: Domain Admins in the domain being prepared. An account with only two of the three fails prerequisite verification.
Is 8418 always just a convergence problem?
No. Microsoft splits it into transient and persistent. Transient ones clear after a replication cycle following a schema update. Persistent ones mean something else is blocking replication – lingering objects, quarantine, DNS or RPC failures, an oversized nTSecurityDescriptor, or database corruption – and that fault has to be found first.
Can I roll the schema back?
No. Schema changes are additive and permanent; classes and attributes cannot be removed once defined. The only direction is forward, which is why the schema master should be backed up before every extension.
I introduced a Windows Server 2022 DC and objectVersion did not change.
That is expected. Windows Server 2019 and 2022 both use schema version 88. A version that does not move is not evidence that preparation failed.
