Hi,
Up until recently, our database servers were running fine. We run SQL 2005 SP4 on Windows Server 2008 R2 SP1. The server has 4x 10 Core CPUs running Hyper-Threaded, so we have 80 cores showing in task manager.
We run 4 SQL instances on an active/passive windows cluster.
All of a sudden (we don't know what triggered it), SQL instances would only run on 40 of the 80 available cores.
After a bit of investigation and reading, we learnt about the Windows hotfix that addressed uneven CPU groups, so we can only presume that we used to be running Group 0 with 64 cores, and Group 1 with 16. We suspect we just never noticed the 16 idle cores.
SQL 2005 is not CPU group aware, so unlike SQL 2008 R2 it will not spread itself across multiple groups.
We found this article: http:/
We tried registry fix in order to put 3 of the NUMA nodes into Group 0, but it simply did not work. We are considering the boot.ini fix, but it makes us nervous, and we've read that that doesnt work either.
We're pretty desperate to get this fix, as our production servers are now under a very high load due to the lack of CPU cores being used.
Does anyone have any experience of this or any tips on what we can do?
Thanks
Russ