One of our clients shut down their db server after a year because it had issues with the domain controller. Since that day the server runs extremly slow. The server has 24cores and 64gb ram What we noticed is an extremly high Latch value in the activity
monitor upto 10000ms. When tracing we saw it came from the tempdb. Until that day it ran on 1mdf, we created 7 additional ndf files for tempdb but no improvement. The latch values stayed high in activity monitor. Then a query i found showed 93pct of resource
wait type to be cxpacket. Decreasing the parallelism to 4 improved the server alot, latch values were gone... Nice workarround, but no real solution. What would be my next step? What would cause this.. A restart of the server a week earlier did not have any
impact on the performance Thanks for reading
↧