PRB: Repl Logreader Fails on Remote Distributor after Cluster Failover

• Microsoft SQL Server version 7.0

SYMPTOMS

When a remote distributor is located on a cluster server, the logreader may fail to run after a failover if the logreader was actively reading at the time of the node failure. The following error occurs:

The process could not execute 'sp_repldone/sp_replcounters' on '<publisher server name>'

CAUSE

An orphaned Server Process ID (SPID) remains on the publisher after the failover. The logreader recognizes this SPID as another logreader on that database and fails to start.

WORKAROUND

Here are two workarounds:

• Identify the orphaned SPID by running sp_who2 on the publisher. The orphaned SPID is the SPID that lists 'Replication-LogReader' as the ProgramName and the publishing database as the database. Use the T-SQL KILL command to clear this orphaned SPID on the publisher, then restart the logreader. Note that if the logreader is still retrying, it will start on its own after you clear the SPID.
  
  -or-



• Modify the Windows NT registry settings to lower the KeepAlive time so that Windows NT will kill the orphaned SPID more quickly, allowing the logreader to resume normal operations. For further information, see the following article in the Microsoft Knowledge Base:

  Q137983 INF: How to Troubleshoot Orphaned Connections in SQL Server

Additional query words: replication

Keywords          : kbSQLServ700 
Version           : winnt:7.0
Platform          : winnt 
Issue type        : kbprb 

Last Reviewed: July 30, 1999