Memory leak in Oracle 8.1.7 ?

[Stefan Farrelly]

Hi all,

Does anyone know of a memory leak in Oracle 8.1.7 ? We have a problem on an N class whereby when all the oracle databases are shutdown each night the amount of user memory does not return to the same level as after a reboot - we are about 1GB short! It appears all shared memory and semaphores are removed aok (ipcs confirms), and all processes stopped, but were still using 1GB more than we should be. Either oracle 8.1.7 isnt freeing up all the memory it was using or we need an hp-ux patch which will free it up. Anyone know of either ?

Cheers,

Stefan

[Steve Steel · Accepted solution]

Hi


try these

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xd3ffa12d6d27d5118fef0090279cd0f9,00.html

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xd2df911284f5d5118ff40090279cd0f9,00.html


steve Steel

[Stefan Farrelly]

Thanks Steve,

very helpful. Definite memory leak in 8.1.7 - will pass to our DBA's/Oracle to get the needed Oracle patches and install.

Cheers,

Stefan

[Tom Geudens]

Hi,
Just a little warning (and this information is "as is") : Do NOT upgrade to the 8.1.7.3 release (8.1.7.2 is fine and will solve your memory problems). We've got very serious problems after this upgrade (with queries / with the system-tablespace / with analyze) and Oracle is still looking for solutions.

Regards,
Tom

[Juan Manuel López]

Could you please give more information about this problem?
We are just beggining to upgrade to 8.1.7.3 from 8.1.6.0. I am very interesting about more information.
Thanks in advance.
Juanma.

[Stefan Farrelly]

Tom,

Thanks for the warning about 8.1.7.3 !

Cheers,

Stefan

[Tom Geudens]

Hi,
The information I have on the 8.1.7.3 problems (2 and 3 are "open calls" with Oracle) :
1) Certain types of complicated queries do NOT work anymore. You can find - so I'm told by our DBA's - information on this at Metalink. This seems to be "a works as designed" problem.
2) Last week we had some processes blocking an analyze. Through searching we located a culprit DBprocess, which on first view seemed blocked itself on an I/O (on the system tablespace), but on second view was CPU looping. We had to rewrite this BO (Bussiness Objects) application.
3) This morning several sqlload's threw a lot of ORA-600 problems (causing the trace directory to "overflow"). This problem is still being looked at, but seems to focus on the system tablespace once more.

Ok, this is the information. It is "as is". These problems may just be "local" problems (well, 1. and 2. are verified to be "global" problems :-). Everything may work fine in your environment. Let's just say that we decided not to continue the upgrades to 8.1.7.3 for the time being.

Regards,
Tom