Memory Leak Problem

Printable View

07-11-2002, 11:04 AM
pareshg

HI,

We are getting memory leak problem while running some online processes.
Platform :Oracle 8.1.7.2/ Solaris.
Somebody told me that it is bug of Oracle 8.1.7.
Is there any solution for this.

Thanks in advance

Paresh
07-11-2002, 11:07 AM
marist89

8.1.7.4
07-11-2002, 11:13 AM
pareshg

Thanks Mariast for ur reply.

That means I need to install patch on my server.
Where will I get it? What is the procedure to install?

Thanks

Paresh
07-11-2002, 11:19 AM
marist89

You get the patches from metalink. There is a readme that will outline the install...
07-11-2002, 11:22 AM
pareshg

Thanks Jeff,

I have one more question for u, that is we will get this memory leak problem for which oracle versions ?

Cheers

Paresh
07-11-2002, 11:29 AM
marist89

I know you get it on 8.1.7.2, that's why I'm upgrading. I don't know about other versions...
07-12-2002, 05:21 AM
julian

Quote:

Originally posted by marist89
I know you get it on 8.1.7.2, that's why I'm upgrading. I don't know about other versions...

There are harmless memory leaks on 8.1.7.3 But Oracle suggested that I "fix" them with event="10262 trace name context forever, level 10000"
07-12-2002, 05:30 AM
gnagesh

Quote:

Originally posted by julian

Quote:

Originally posted by marist89
I know you get it on 8.1.7.2, that's why I'm upgrading. I don't know about other versions...

There are harmless memory leaks on 8.1.7.3 But Oracle suggested that I "fix" them with event="10262 trace name context forever, level 10000"

I believe it is not enough, we need to set _db_handles_cached=0

Right!
07-12-2002, 05:34 AM
julian

For me it was enough.
07-12-2002, 05:37 AM
gnagesh

I keep _db_handles_cached=0

Is there any harm? Please correct me.
07-12-2002, 05:42 AM
julian

Quote:

Originally posted by gnagesh
I keep _db_handles_cached=0

Is there any harm? Please correct me.

I think yes, because with _db_handles_cached =0 you are actually setting off the buffer handles, i.e, you have no handles cached per process. Thus your performance my suffer.
07-12-2002, 06:02 AM
gnagesh

Quote:

Originally posted by julian

I think yes, because with _db_handles_cached =0 you are actually setting off the buffer handles, i.e, you have no handles cached per process. Thus your performance my suffer.
[/B]

Very good input. So you mean to say that "event="10262 trace name context forever, level 10000" solves the memory leak problem. Why I am asking this, I had this problem, and I set both the parameters. It is Ok now. Since _db_handles_cached=0 hurts the performance, no meaning to set this parameter. Some times exploring too many documents also creates problems? am I correct?

Thanks In Advance.
07-12-2002, 06:12 AM
julian

Yes, you are right. 10000 means that you will not get error messages for leaks less than 10000 bytes. They are usually at most 5000. But I was told by Oracle that (was it) 0 would neglect all leaks.