Over the past week and a half we started experiencing a sporadic slowdown in our production x64 SQL 2005 Ent. Edition server. Users started complaining of slowness then they started getting timeouts. In looking at sp_who2 and perfmon we saw the following during the slow/frozen periods:
* Dramatic increase in Perfmon Active Transactions
* CPU higher than norm, but not dramatically so
* sp_who2 shows a number of spids in SUSPENDED state (and not running waits)
* no blocking indicated from sp_who2
* active connections slowly increasing
* no disk queuing (or at most some spikes to 1)
After a couple of minutes of this we would then see the following:
* no more spids in SUSPENDED state
* Logins per second spikes dramatically
* Active transactions spikes down to "normal levels"
* CPU goes high then levels out at moderately higher than normal
* active connections slowly decreases back towards normal levels
* large spike in lock wait time
We turned on the Async Auto Update Statistics option (after testing in our staging environment) on the primary database about a week before we saw this problem. By turning it off we can visually see the problem go away by watching the above metrics. So my question is, What metrics can I use to see the "blocking" or resouce locking that is causing these problems
Anyone
Thx
Ron