I recently spent several days investigating a replication failure. Finding the answer specific to my problem was difficult, so I thought I would post my findings and hope someone else may find it useful.

The day after installing a few security patches, replication to the subscriber database began to fail. The Distribution Agent had the following error message:

This message is usually caused by changing the SQL Server service using the Windows Service Control manager, which cannot grant the required permissions to start the service appropriately or to run SQL Agent jobs. SQL Server Configuration Manager should be used instead and the correct way to fix it is to set the service account to the Local System account, then back to the domain account using the SQL Server Configuration Manager. (http://support.microsoft.com/kb/911305/)

Since I did not change the service account recently, I thought this could not be the correct answer to the problem. But since I did get the same error message, I thought it would not hurt to restart the service with SQL Server Configuration Manger anyway. Of course, this did not work, but error message does mean it was security related.

Then I changed all the replication agents to have the maximum privileges possible. This did work either. This led me to think that this was not related to the articles themselves and maybe not even to Replication or SQL Server. I found the following error message in Event Viewer:

This error message occurs when two or more computer accounts have the same SPN registered.
(http://support.microsoft.com/kb/321044)

To fix this we deleted the computer account entries in AD, then disjoined and rejoined the server from the domain. Except for a few expired articles, all replication articles synchronized on their own without intervention. I reinitialized the expired articles.

This fixed our problem so far, but if anyone has any experience with this your feedback is welcomed.

Regards,

Jon