Re: PGA increasing continuously

From: Purav Chovatia <puravc_at_gmail.com>
Date: Sat, 5 Dec 2020 01:27:31 +0530
Message-ID: <CADrzpjFzmYGKNyaeKUJjFkWo1tpuhgJbbOFptoLJxNcu_P_Cww_at_mail.gmail.com>



Thanks. We applied the Oct 2020 RU, so I guess it should have got included, but checking...

On Sat, 5 Dec 2020 at 00:37, Laurentiu Oprea <laurentiu.oprea06_at_gmail.com> wrote:

> Did you check if you hit Bug 17871069?
>
> În vin., 4 dec. 2020 la 19:36, Stefan Koehler <contact_at_soocs.de> a scris:
>
>> Hello Purav,
>> „kxs-heap-p“ is just the (sub-)heap. What are the allocated chunks about
>> inside this (sub-)heap?
>>
>> As you mentioned you are running on Solaris, you can use a DTrace script
>> of mine to identify the code path that allocates these memory chunks:
>> http://soocs.de/public/scripts/dtrace_kghal_pga_code
>>
>> Hope this helps.
>>
>> Best Regards
>> Stefan Koehler
>>
>> Independent Oracle performance consultant and researcher
>> Website: http://www.soocs.de
>> Twitter: _at_OracleSK
>>
>> > Purav Chovatia <puravc_at_gmail.com> hat am 04.12.2020 17:01 geschrieben:
>> >
>> >
>> > Hello Experts,
>> >
>> > Need some help desperately.
>> >
>> > We have an application that is doing a StoredProc call and a SQL call
>> from a Standalone Java App which uses the Thin driver.
>> >
>> > The StoredProc does an UPDATE with a WHERE clause on the single column
>> Primary Key.
>> > If the SQL%ROWCOUNT=0 then it will do an INSERT.
>> >
>> > The SQL does a select on the table with a WHERE clause on the Primary
>> Key.
>> >
>> > No joins. No CLOB/BLOB/LOB. No XML. Prepared Statements used. Only
>> thing that stands out is that the table has 100+ columns.
>> >
>> > PGA goes on increasing and hence free memory on DB server goes on
>> decreasing.
>> >
>> > Problem originally started with ojdbc7.jar from 12.1.0.1.
>> > We tried the ojdbc7.jar from 12.1.0.2 and the problem worsens.
>> > Problem persists with ojdbc6.jar from 11.2.
>> > With ojdbc6.jar problem does not reproduce.
>> >
>> > Problem persists with DB 11.2.0.4, 12.1.0.2. Problem persists on DB
>> server on Linux as well on Solaris.
>> >
>> > On DB server, we dont see any open cursor problem. No error on App side
>> or on DB side except ORA-04036 in case of 12c because it will cap PGA with
>> pga_aggregate_limit.
>> > We dug on PGA front and identified that subheap kxs-heap-pis where the
>> memory is used. But cannot understand what does that mean?
>> >
>> > Would appreciate any ideas.
>> >
>> > It was a shock to realize that there is no way beyond identifying the
>> subheap! Developers are ready to fix the issue but they have been
>> requesting to tell what is to be fixed. Is there no way to identify what is
>> eating up the PGA?
>> >
>> > Thank you.
>> --
>> http://www.freelists.org/webpage/oracle-l
>>
>>
>>

--
http://www.freelists.org/webpage/oracle-l
Received on Fri Dec 04 2020 - 20:57:31 CET

Original text of this message