Re: Running utlrp creates too many Pnnn processes

  • From: Yong Huang <yong321@xxxxxxxxx>
  • To: oracle-l@xxxxxxxxxxxxx
  • Date: Fri, 19 Dec 2008 06:20:56 -0800 (PST)

See Bug 5898416 and Data Warehousing Guide. Parallel_max_servers is calculated 
based on some unpublished formula. In 10g, I often see it at 20 times cpu_count 
but it varies. I can't think of a good reason why parallel_max_servers needs to 
be set that high. Just explicitly set it to a lower value.

(BTW, I'm writing a set of programs to determine what parameters affect what 
others. So far I've only finished session-modifiable boolean and numeric type 
parameters.)

Yong Huang

> That solution was suggested by another Oracle-L member, too (thanks Paul).
> 
> In the meantime, I did manage to sort of determine the reason for the
> problem.
> 
> The cpu_count parameter is 64.  It is one of the parameters used to
> calculate the value of the parallel_max_servers parameter.  The value
> of the processes parameter is also used in that calculation, though it
> is not mentioned in the Reference Guide.  When I increased the value
> of processes, the value of parallel_max_servers increased.  I only
> did a few tests and found that at processes=300, parallel_max_servers
> was set to 285, but when processes was set to 1500, parallel_max_servers
> was only 1280 and I only had 276 parallel query slaves start up.
> 
> Unfortunately, we don't have enough memory available to run the database
> with processes set to 1500....  So, one solution is to set the
> parallel_max_servers parameter to something lower than the calculation,
> or run ultprp.sql with a parameter instead of running ultrp.sql.
> 
> - Maureen
...
>> successfully and patching to 10.2.0.3.0 also completed successfully.


      
--
//www.freelists.org/webpage/oracle-l


Other related posts: