Re: Running utlrp creates too many Pnnn processes
Date: Mon, 22 Dec 2008 09:02:50 -0900
That is exactly what our problem is! I spent all day searching through Metalink articles and put in a ticket with Oracle, too. Even after I sent them the alert log and numerous trace files, they asked that I send the trace file from when the error occurred...which is what I had sent....
I'll pursue changing the cpu_count parameter because, as you said, it will solve the majority of the problems.
We, too, have numerous memory issues on these machines...I'm not sure exactly what they are, but they're Solaris SPARC 64-bit machines, using Sun Clusters. There are certainly a lot of them, but only 4 that I'm dealing with directly.
Thanks for the info!
Finn Jorgensen wrote:
> Are you on a T5220 or T5240 or something like that by any chance with
> highly threaded CPU's? These servers have many problems related to
> CPU_COUNT being set high. We have a T5240 with 128 virtual CPU's and ran
> into many different issues related to memory and this parameter. I have
> also run into your exact problem on a T5220 with 64 virtual CPU's.
> Anyway we have so many of these servers now I finally opened a SR with
> Oracle and asked what the implications of lowering the CPU_COUNT
> parameter was (because Oracle says NOT to touch this parameter) and the
> answer was that it's fine to lower the parameter. This will solve many
> of your problems. Just set it to 32 or 16 or whatever works for you.
> On Thu, Dec 18, 2008 at 4:53 PM, Maureen English <sxmte_at_email.alaska.edu
> <mailto:sxmte_at_email.alaska.edu>> wrote:
> 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
> - Maureen