Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Response and performance issue

Re: Response and performance issue

From: kyle hailey <oraperf_at_yahoo.com>
Date: Thu, 22 Feb 2001 01:51:42 -0800
Message-ID: <ism99tcmk1nrb988rg2u4pkqiv9jtttle8@4ax.com>

Whoops, I think I got your mail mixed up with another similiar one.

For this case, the easiest think might be to run (this is for sqlplus)

   REM you need timed statistics to get much out of the following    alter session set events='10046 trace name context, level 8'    REM you need to have run @?/rdbms/admin/utlxplan for the following    set autotrace on

This gives you

  1. wait events in the trace file
  2. session statistics as part of the output

The audit triggers I mentioned before do both of these things but are a little work to set up. The one advantage is that the audit triggers put the info into tables that are easy to analyze.

Writing a script to sum the waits in the trace file is pretty easy. Here is and example:

cat $* |\

            grep WAIT |\
            sed -e 's/ela= \([0-9]*\) /,\1,/g' |\
            sed -e "s/'/,/g" |\
            awk -F, '
             BEGIN{wait[0]=0;elap[0]=0}
            {
              wait[$2]++
              elap[$2]=elap[$2]+$4
            }
            END{ for (i in wait) printf("%30s %10d %12d\n" ,
i,wait[i],elap[i])}'

Curious to know what the difference is.

If the stats and the wait events match you might want to try tracing at the oracle net level and as I mentioned trussing.

Best
Kyle Received on Thu Feb 22 2001 - 03:51:42 CST

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US