How much overhead will 10046 trace carry when it's enabled?
Summary:
Platform: Linux, Oracle Database 11g Enterprise Edition Release 11.2.0.4.0 - 64bit Production.
A report started to run late on my production database,
- I noticed in ASH, one particular insert SQL_ID appeared around 70% in the overall sample (per second),
- When I checked SQLSTATS for the insert, elapsed time was significantly lower than the number of samples reported (during the cycle),
- Then I analyzed the ASH data and noticed - samples are not continuous (difference between two samples is also random, varied between 02 and 05 seconds),
- I checked CPU utilization at the time - CPU was 60% idle, so it's not due to system resources busy during that time,