Re: kolccst obj: large streams pool memory segment

We are still dealing with this issue, and still curious about a workaround. Here are some interesting facts from the one database exhibiting this problem:

STRMDEV_SQL > show sga

Total System Global Area 943718400 bytes

Fixed Size                  2033448 bytes
Variable Size             515905752 bytes
Database Buffers          419430400 bytes
Redo Buffers                6348800 bytes

STRMDEV_SQL > select pool, sum(bytes)/1024/1024 from v$sgastat group by rollup(pool);

POOL SUM(BYTES)/1024/1024

------------ --------------------
java pool                      24
shared pool            152.007301
streams pool           1660.38328
                       407.993935
                       2244.38451

STRMDEV_SQL > show parameter sga

NAME                                 TYPE        VALUE
------------------------------------ --------
------------------------------
lock_sga                             boolean     FALSE
pre_page_sga                         boolean     FALSE
sga_max_size                         big integer 900M
sga_target                           big integer 0

STRMDEV_SQL > col value for 999,999,999,999 STRMDEV_SQL > select * from v$sga;

NAME                                                VALUE
---------------------------------------- ----------------
Fixed Size                                      2,033,448
Variable Size                                 515,905,752
Database Buffers                              419,430,400
Redo Buffers                                    6,348,800

How is it that the Streams Pool can grow unbounded? I thought it came out of the buffer cache when sga_target is unset (as documented<http://download.oracle.com/docs/cd/B19306_01/server.102/b14220/memory.htm#sthref1394>). Or is that part of the bug as well?

And just for all that good background stuff: OSEE 10.2.0.2 on Solaris 8

On 10/4/07, Charles Schultz <sacrophyte_at_gmail.com> wrote:
>
> Any workaround? I am curious if turning off the evil ASMM will bypass the
> issue or not. Surely we can set hard memory limits, but does that have other
> ramifications for this Streams-specific bug?
>
> On 10/4/07, Rajeev Prabhakar <rprabha01_at_gmail.com> wrote:
> >
> > Charles
> >
> > From what I could find, it looks like the "kolccst obj" are used to
> > maintain the streams heap collection.
> >
> > Rajeev
> >
>

-- 
Charles Schultz

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