Fix it now
Error 8344 is Active Directory reporting ERROR_DS_INSUFF_ACCESS_RIGHTS, insufficient access rights to perform the operation. The account Entra Connect uses against your directory does not have the permission the operation needed, or the server can no longer reach the cloud endpoints at all.
Get-Service ADSync
Get-ADSyncScheduler
Get-ADSyncADConnectorAccount
Invoke-WebRequest -Uri https://adminwebservice.microsoftonline.com/ProvisioningService.svc
- Read the scheduler output for the sync cycle being enabled and for staging mode. A staging server produces runs that complete and change nothing.
- Read which account the AD connector is using, then confirm that account’s password is current. If it changed, re-enter it in Synchronization Service Manager under Connectors, the Active Directory Domain Services connector, Properties.
- Read the result of the web request. HTTP 200 is a working path. A 403 means the proxy has not been opened for that URL; a 407 means the proxy wants authentication and none was supplied.
- Re-apply the rights for the features you actually use with the documented ADSyncConfig cmdlets –
Set-ADSyncBasicReadPermissionsfirst, then only the ones matching your enabled features. - Restart the service and force a cycle:
Restart-Service ADSyncthenStart-ADSyncSyncCycle -PolicyType Delta.
Grant only the rights the features you use require. Making the connector a domain administrator will clear the error and turn a compromise of this server into a compromise of the whole directory.
If the cycle completes and the failing operation works end to end, stop here. If only one feature or one organisational unit fails, the next section maps each feature to the right it needs.
Why it happens
Directory synchronisation uses more than one identity and they fail differently. The service account runs the synchronisation service and holds its configuration database. The AD DS connector account reads your on-premises directory and, depending on which features are enabled, writes back to it. A third account exists in the cloud and is created and maintained by the configuration wizard. Error 8344 points at the second of these almost every time, because it is the only one that has to satisfy Active Directory’s access control – 8344 is ERROR_DS_INSUFF_ACCESS_RIGHTS, insufficient access rights to perform the operation.
What the connector account needs depends on which features are switched on, and Microsoft documents a cmdlet per feature. Basic read needs read on all properties of descendant objects plus Replicate Directory Changes at the domain root. Password hash synchronisation needs Replicate Directory Changes and Replicate Directory Changes All at the domain root. Password writeback needs Reset Password and write on lockoutTime and pwdLastSet. Group writeback, Exchange hybrid, Exchange mail public folders and mS-DS-ConsistencyGuid each have their own. Turn a feature on without granting its rights and only that feature fails, which is why partial failures are so informative.
Those rights also drift. Inheritance disabled on an organisational unit stops delegated permissions reaching the objects inside it – Microsoft ships Get-ADSyncObjectsWithInheritanceDisabled precisely to find that. A hardening exercise can strip delegations nobody documented, and accounts protected by the directory’s own privileged-account mechanism have their permissions periodically reset, so a delegation applied to them does not stay applied.
The connector account’s password changed or expired
You have this one if Every operation against the on-premises directory fails at once, starting at a time somebody was doing account maintenance.
- Confirm which account is in use with
Get-ADSyncADConnectorAccount. - Set a known password on it, then enter that password in the connector properties in Synchronization Service Manager.
- Restart the ADSync service and run a delta cycle.
- Configure the account so its password does not expire, and record where it is documented.
Delegated permissions were removed or inheritance was disabled
You have this one if One feature, or one organisational unit, fails while the rest of the synchronisation works normally.
- Run
Get-ADSyncObjectsWithInheritanceDisabledto find containers where inheritance is blocking the delegation. - Re-enable inheritance where it was turned off unintentionally.
- Re-apply the rights with the ADSyncConfig cmdlet for the feature that is failing, scoped to the connector account.
- Use
Show-ADSyncADObjectPermissionsto confirm the rights actually landed on the object.
Run those cmdlets from an elevated session on the synchronisation server, signed in as an account that can change directory permissions.
The server cannot reach the cloud service
You have this one if Every object fails at the same moment, and the failure references a cloud endpoint rather than a directory object.
- Run the documented connectivity test and read the status: 200 is healthy, 403 means the URL is not opened on the proxy, 407 means the proxy requires authentication that was not configured.
- Configure proxy settings in machine.config, which is what the wizard and the sync engine read – not miiserver.exe.config, which Microsoft warns is overwritten on every upgrade.
- Confirm the proxy idle timeout is greater than 5 minutes, which Microsoft documents as a cause of intermittent connectivity.
- Restart the ADSync service after any machine.config change.
The cloud side of the connection needs re-authorising
You have this one if On-premises operations succeed, the cloud connector fails, and nothing about the network changed.
- Re-run the configuration wizard and re-enter credentials for an account holding the hybrid identity administrator role.
- Let the wizard recreate or refresh the cloud synchronisation account; it is not an account to edit by hand.
- Confirm the account has not been caught by a Conditional Access policy applied to all users, which is a common accident.
The server is in staging mode or the scheduler is off
You have this one if Runs complete without errors and nothing ever changes in the cloud.
- Read
Get-ADSyncSchedulerfor the sync cycle being enabled and for staging mode. - Re-enable the cycle if it was paused deliberately and never resumed.
- Confirm which server is the active one before changing staging mode on either.
Full reference
One cmdlet per feature
| Feature | Rights it needs | Cmdlet |
|---|---|---|
| Basic read | Read all properties on descendant objects; Replicate Directory Changes at the domain root | Set-ADSyncBasicReadPermissions |
| Password hash synchronisation | Replicate Directory Changes and Replicate Directory Changes All, domain root only | Set-ADSyncPasswordHashSyncPermissions |
| Password writeback | Reset Password; write lockoutTime and pwdLastSet on descendant users | Set-ADSyncPasswordWritebackPermissions |
| Group writeback | Generic read/write, create and delete child objects, delete tree on group objects | Set-ADSyncUnifiedGroupWritebackPermissions |
| Exchange hybrid | Read and write all properties on users, InetOrgPerson, groups and contacts | Set-ADSyncExchangeHybridPermissions |
| Exchange mail public folders | Read all properties on descendant PublicFolder objects | Set-ADSyncExchangeMailPublicFolderPermissions |
| mS-DS-ConsistencyGuid | Read and write that property on descendant user objects | Set-ADSyncMsDsConsistencyGuidPermissions |
Set-ADSyncRestrictedPermissions goes the other way, tightening permissions on the connector account itself. Apply the feature cmdlets for what you use and nothing else; the point of having seven of them is that a synchronisation server does not need one broad grant.
What is failing tells you which right is missing
| What fails | Where to look |
|---|---|
| Only password hash synchronisation | Replicate Directory Changes and Replicate Directory Changes All at the domain root |
| Only writeback operations | Write permissions on the target attributes or organisational unit |
| Everything in one organisational unit | Inheritance disabled on that container |
| Only administrative accounts | Protected-account permission resets have removed the delegation |
| Nothing synchronises and the failure names a cloud endpoint | Outbound connectivity, the proxy, or the cloud account |
| No errors but nothing exports | Staging mode, or the scheduler is disabled |
Proxy configuration, done the supported way
The installation wizard and the sync engine are .NET applications, so Microsoft requires the proxy to be configured in machine.config. There is a long-standing habit of editing miiserver.exe.config instead, which Microsoft addresses directly: that file is overwritten on every upgrade, so even where it works initially the system stops working at the first upgrade. Configure machine.config, restart the ADSync service, and re-run the connectivity test.
- A 200 response from the provisioning service URL means the path works.
- A 403 means the proxy has not been opened for that URL – revisit which URLs are permitted.
- A 407 means the proxy required a sign-in and none was provided; if your proxy authenticates, that has to be configured in machine.config.
- A connection error means the call went direct, or name resolution failed on the server itself.
- A proxy idle timeout of 5 minutes or less causes intermittent failures that look like something worse.
On the three event IDs
Event ID 6801, Event ID 109 and Event ID 115 are commonly quoted with this error, and Microsoft publishes no meaning for any of them in this context. They are not a diagnosis and this article will not invent one. The run profile results in Synchronization Service Manager, and Microsoft Entra Connect Health, both name the connector, the run step and the object that failed, which is what you actually need.
Never take two synchronisation servers out of staging mode at the same time. Two active servers writing to the same tenant will fight over objects, and untangling that is considerably harder than the outage you were trying to fix.
Every code this article covers
| Code | What it points at | Source |
|---|---|---|
8344 |
ERROR_DS_INSUFF_ACCESS_RIGHTS (0x2098): insufficient access rights to perform the operation | Microsoft Learn |
Event ID 6801 |
Quoted alongside failed synchronisation runs. Microsoft publishes no meaning for this event ID; read the run profile results in Synchronization Service Manager instead | not published by the vendor |
Event ID 109 |
Quoted around synchronisation failures. Microsoft publishes no meaning for this event ID in this context | not published by the vendor |
Event ID 115 |
Quoted alongside the same failures. Microsoft publishes no meaning for this event ID in this context | not published by the vendor |
Confirm the fix worked
- A delta cycle completes with no export errors on any connector.
- The specific operation that was failing – a password change flowing through, or a writeback – now works end to end.
Invoke-WebRequestagainst the provisioning service URL returns 200 from the synchronisation server itself.Get-ADSyncSchedulershows the cycle enabled and staging mode set the way you intend, on both servers if you have two.
Questions people ask about this
Does fixing this cost anything?
No. Directory synchronisation is included with the subscriptions that use it, and everything here is permissions, credentials and network configuration.
Can I just use a domain administrator account for the connector?
It will work and you should not. The connector account is stored on a server that talks to the internet, and giving it full directory rights turns a compromise of that server into a compromise of the whole directory. Microsoft ships a cmdlet per feature so you do not have to.
Why does only password hash sync fail?
Because it needs rights nothing else does: Replicate Directory Changes and Replicate Directory Changes All at the domain root. Set-ADSyncPasswordHashSyncPermissions applies them.
Should I edit miiserver.exe.config for the proxy?
No. Microsoft says plainly that it is overwritten on every upgrade, so the fix survives until the next update and then fails. The supported location is machine.config, followed by a restart of the ADSync service.
How long can synchronisation be down before users notice?
Existing accounts keep working, because the cloud copy is already there. What stops is change: new starters do not appear, leavers stay enabled, group membership drifts, and with password hash sync down, on-premises password changes never reach the cloud.
