The issue below took me a good 6 hours to fix. Hopefully it will save somebody that time, so here it is:
My Tridion 2011 SP1 HR1 Content Manager was getting hit by a very peculiar performance issue – when saving, creating or deleting any item, the operation would be very slow (5+ seconds). This was on a controlled environment (my VMWare) with nobody else on the system. The issue was manifesting in both the CME and with API (Core Service and TOM.NET).
After checking the usual culprits (disable Event Systems, disable CME extensions, execute sp_updatestats, rebuild all tables index, repair SDL Tridion CM, uninstall/install CM), I finally noticed in the Tridion event log the following line (in fact there were hundreds of these warnings):
Unable to notify "T2011SP1". Reason: The requested name is valid, but no data of the requested type was found
Component: Tridion.ContentManager
Errorcode: 0
User: T2011SP1HR1\MTSUser
The host T2011SP1 is the old hostname of my VMWare. I make a habit that I rename the servers names to reflect the version of Tridion I’m running. This immediately led me to go to my Tridion_cm database and delete the entries from QUEUE_CONSUMERS and QUEUE_FILTERS tables that have HOST = ‘T2011SP1’. A less drastic measure would have been to simply set their respective column IS_ONLINE = 0.
Restarted my Tridion * services, shutdown the Tridion COM+ application and magically performance was back as it used to. Now I can batch update existing items via the Core Service in about ~0.25s per item.
What seemed very strange was that the ‘write’ operations were being notified to the other CM instances present in the QUEUE_CONSUMERS table. That table contains messages of type Publish, Deployer, Workflow, and Search. I guess it was the Search that was being updated… not sure.
Comments