Re: Connection spike resmgr:internal state change
- From: Mladen Gogala <gogala.mladen@xxxxxxxxx>
- To: oracle-l@xxxxxxxxxxxxx
- Date: Sun, 4 Sep 2022 13:44:43 -0400
On 9/4/22 02:00, Andy Sayer wrote:
Try gv$event_histogram rather than sampled ASH.
When we raised the SR 3 years ago we were still on 12.2. The issue
still exists, there’s been no updates as far as I know (of course we
had to close the SR because the ER had been logged).
If you are not really using resource manager, you could just unset the
resource plan (it is useful during evening and weekend maintenance but
that is when it shouldn’t matter too much). I decided to just live
with the 0.1 second spikes and let Oracle fix it in their own time. As
long as you have sensible connection pools, a 0.1 second pause in all
calls shouldn’t be too noticeable to your users.
Hi Andy,
100 milliseconds every 20 minutes doesn't look like a big deal. It looks
like Oracle is emulating the operating system, with that wait playing
the role of the clock interrupt. Everything has to wait for the clock
interrupt to complete and statistics is gathered during the clock
interrupt processing. Also, scheduler is activated and process
scheduling is done: signal delivery, quantum expiration etc.
Oracle needs to do the same, in order to limit CPU processing and memory
consumption. It needs to make everything stop and check the state. As
the great bard once wrote:
But thoughts the slave of life, and life, Time’s fool,
And Time, that takes survey of all the world,
Must have a stop.
Oracle RDBMS too must have a stop every once in a while.
PS:
----
The quote above is not from Oracle documentation, although Oracle
documentation frequently uses Shakespeare's style, apparently to be less
understandable. To process or not to process, that's the question now.
The quality of the documentation is extremely low.
--
Mladen Gogala
Database Consultant
Tel: (347) 321-1217
https://dbwhisperer.wordpress.com
Other related posts:
