Comments on: SESSIONS Derived Value Formula – Doc Bug Explains

By: Charles Hooper

Charles Hooper — Thu, 25 Apr 2013 11:10:28 +0000

PC,

This morning I tried digging through My Oracle Support (MOS) to see if there had been any discussion regarding the derived value of SESSIONS (I would like to mention that the search functionality is very limited in MOS – many unrelated articles are returned). This time I managed to find a bug report titled “Bug 9357816 : [DOCBUG] DEFAULT VALUE OF SESSIONS IS CHANGED ABOVE 11.2″ that was last updated late in February 2010. A user reported several pairs of PROCESSES and derived SESSIONS pairs, and in each case the derived value of SESSIONS could be determined based on the following formula:

(1.5 * PROCESSES) + 22

The documentation page that I linked to in this article is no longer found on Oracle’s website. So, I searched through the documentation and found that the page describing the SESSIONS parameter was updated sometime after my article was posted. The new documentation reference is found here:
http://docs.oracle.com/cd/E11882_01/server.112/e25513/initparams232.htm#REFRN10197

The documentation now shows:
Derived: (1.5 * PROCESSES) + 22

While the documentation is closer to being correct now, the new formula in the documentation does not agree with the formula that anyone commenting on this blog article has provided. Of the 5 sample PROCESSES and SESSIONS pairs found in the table of this article, the documentation’s new formula only works for 1 of the PROCESSES and SESSIONS pairs.

I located a couple of 10+ year old articles in MOS that suggested that enabling auditing may cause the formula for deriving the SESSIONS parameter value, but I do not know if that has any impact.

By: PC

PC — Thu, 25 Apr 2013 09:28:29 +0000

On Solaris x86_64, Oracle SE 11.2.0.2, with processes=500, sessions was 792. So that works out to processes * 1.5 + 42. On another Solaris x86_64, Oracle SE 11.2.0.2.8, with processes=2000, sessions was 3024. So that works out to processes * 1.5 + 24. (Value modified at the request of PC, CH April 25, 2013)

By: Charles Hooper

Charles Hooper — Wed, 06 Jun 2012 01:16:20 +0000

CVDV,

Thanks for confirming that the change in the behavior started with 11.1.0.6.

By: CVDV

CVDV — Tue, 05 Jun 2012 23:55:37 +0000

And I have confirmed that the behaviour of 11.1.0.6 is different again. Setting processes to 300 and bouncing the instance yielded sessions of 335 and transactions of 368. This conforms to the published Oracle documentation formulas (sessions=processes * 1.1 + 5 and transactions=sessions * 1.1). so it would appear that the formulas changed in 11.2 and that the documentation was never updated.

By: CVDV

CVDV — Tue, 05 Jun 2012 04:41:31 +0000

I can confirm for 11.2.0.3. I changed processes to 300 and sessions were automatically set to 472 after bouncing the instance. If I get around to testing against 11.1.0.6 I’ll bump those results here too.
Thanks for the excellent post.

By: Log Buffer #258, A Carnival of the Vanities for DBAs | The Pythian Blog

Log Buffer #258, A Carnival of the Vanities for DBAs | The Pythian Blog — Fri, 10 Feb 2012 09:56:13 +0000

[...] Charles Hooper explains a Doc Bug. [...]

By: Charles Hooper

Charles Hooper — Wed, 08 Feb 2012 11:20:55 +0000

Timur,

I considered submitting a comment in the documentation, but I have not done so. I am curious to know:
* Does the new formula apply to 11.2.0.2 on all operating system platforms?
* Does the new formula apply to 11.2.0.1 and 11.2.0.3, and what about 11.1.0.6 and 11.1.0.7?
* 11.2.0.1 was released September 1, 2009 – has anyone else reported the apparent error in the documentation?

On a side note, the non-default CURSOR_SHARING parameter values are still having issues in 11.2:
http://afatkulin.blogspot.com/2012/02/sessioncachedcursors-again.html

By: Timur Akhmadeev

Timur Akhmadeev — Wed, 08 Feb 2012 06:54:18 +0000

Charles

have you posted a comment in the docs about this behavior?