Yes, running vmstat almost always shows me an overloaded runable vs blocked queue. Much of the time we are able to reel the monster in by finding very long running queries, blocked sessions, or duplicate/runaway process (all application problems). This of course requires that someone sit and watch for the condition to occur essentially from beginning to end of business day, or wait for users to call up and say the system has become unusable. Neither acceptable. Memory is not a problem and io appears to be well within the limits for nearly 100% of the time. Once in a while some report or batch process will tax a drive but that case is rare. I've found that throwing hardware resources at software issues almost always results in the problem just being deferred until later down the road; the fix is only temporary. Bad sql can always end up dragging down the best hardware that can be thrown at it.

Mike

>>> "Kirt Thomas " <oracledba-ezmlmshield-x19809964.[Email address protected] 5/31/2006 9:48:51 AM >>>
Don't forget estat/bstat for the db, and vmstat for the os. Topas
should be able to tell you especially if there is a memory
issue. The only times I've seen AIX choke on an Oracle DB were when
there was an i/o bottleneck (ie exceeding the physical i/o possible
on the hdisks) or when memory started paging. Both of these can be
seen in topas. I mainly use vmstat to look at runnable vs blocked processes.




Oracle LazyDBA home page