Re: cpu waits

Alex

If you are on *nix, you should investigate basic OS performance monitoring 
utilities, eg:

sar
iostat
vmstat
top (if available)
(there are others...)

For example, sar -q  will report the run-queue size; see if your SA will (or 
has) set up a sar database. At my last site we recorded sar output on a minute 
by minute basis through the day, and kept 8 days history - so it's always 
possible to go back and look at what just happened, or to compare with the same 
time yesterday. The overhead is fairly negligible (and because it is always on, 
you never have to worry about forgetting to kick it off).  Manual pages and 
Google are your friends - there are lots of articles and books out there.

I blogged a while ago on how to record and load iostat output into a database 
for offline analysis (or correlation with statspack reports) here: 
http://preferisco.blogspot.com/2006/09/loading-iostat-output-into-oracle.html. 
You can use the same approach for vmstat etc. You could even pipe iostat output 
into SQL*Loader, and have near real time delivery of the data into your 
PERFSTAT schema...

You mostly can't cross reference stats to specific processes - but you should 
be able to drill down onto strange occurrences by time. It's better than 
nothing!

Good luck

Regards Nigel
----- Original Message ----
From: amonte <ax.mount@xxxxxxxxx>

I was wondering what event does Oracle show when the process cannot get CPU 
cycles?

TIA

Alex

Other related posts: