Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: 9.0.2 64-bit on redhat 3.0r2 AMD64 (4-way) 17G RAM

Re: 9.0.2 64-bit on redhat 3.0r2 AMD64 (4-way) 17G RAM

From: Paul Drake <bdbafh_at_gmail.com>
Date: 27 Aug 2004 21:29:03 -0700
Message-ID: <910046b4.0408272029.1cfbf4b3@posting.google.com>


wealtheow1_at_yahoo.com (Artis Gripemore) wrote in message news:<69750f44.0408271325.74110a14_at_posting.google.com>...
> Please ignore this if the questions of amatuers are not allowed here.
> I thought that perhaps the rare hardware configuration might be of
> interest, however.

damn straight.

> Thanks in advance.
>
> I have shmmax set to 8G, but I cannot get db-cache_size over 2G or
> oracle runs out of memory. What have I done wrong?
>
> using--
>
> db_cache_size at just under 2G (and it cannot be increased without
> failure)
> pga_aggregate_target at 4G
> workarea_size_policy set to auto
> shared_pool_size is 24M
> shared_pool_reserved_size is 2M
>
> block size is 2048
>
> (the other settings are quite conservative, but feel free to ask)
>
> I am sure that I am doing something wrong, but I've tried a bunch of
> different sysctl.conf variations and I still get trouble increasing
> the buffers. If anyone would like to suggest a sysctl.conf
> appropriate to the hardware, or show me the error in this init.ora, I
> would appreciate it. Please ask if you need more info.
>
> Thanks again.
>
> S

drool ...

posters to oracle-l have reported issues with pga_aggregate_target over 1GB.
what is your value for sga_max_size?

how about getting it to work past the 2 GB hurdle, then 4 GB, then go for however big you have determined is necessary.

you don't need to fit the entire sga into a single shared memory extent, but its one way of doing it.

it it possible that you are asking for an 8 GB shared memory size, but its only actually being set at 2 GB? In this case, might you be able to get the db_cache_size in one 2 GB segment, keep + recycle in another, other block size caches in another ... I don't know how that would actually get co-ordinated, but its a thought.

have you considered abandoning pga_aggregate_target in favor of pga/uga being under each dedicated server process?

btw1 - why don't you throw 128 MB at the shared pool for a start?

btw2 - are you a fan of the "A" train?

-bdbafh Received on Fri Aug 27 2004 - 23:29:03 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US