Re: Memory leak

This is classical Unix memory management FAQ.

A good operating system (unlike DOS for example) will use all available memory, because any unused memory is wasted memory. Memory is used for file system caching for example, it's still better than wasting memory completely by not using it.

You should check paging activity using sar -d and possibly page scan rate using sar -g or vmstat. If these are constantly high, you have a problem. If not, you have no memory shortage. If you ask, what is "high", then you have to do research about it on your own. Generally, when I see few hundred kilobytes of sustained IO activity on swap disks, I don't count it as a real problem. If we are talking several megabytes per sec and high page scan rates, then we are dealing with memory shortage problem.

Tanel.

• Original Message ----- From: "Adolfo" <air_at_tid.es> Newsgroups: comp.databases.oracle.server Sent: Tuesday, September 23, 2003 4:56 PM Subject: Re: Memory leak

> We check with top and sar -r
>
> Our shared_pool_size is 300000000 and our RAM is 3Gb. Someone is eating
up the
> memory until we only have 40 Mb of the 3Gb. I don't undertand why oracle
takes
> more memory than its 300Mb
> When we stop the select and shutdown oracle we only recover about 300 Mb
and
> with /usr/ucb/ps -aux we don't see which process have the memory
>
> Adolfo
>
> Tanel Poder ha escrito:
>
> > How do you see that system doesn't recover the memory? Do you get any
errors
> > or just checking top output?
> >
> > It's probably not memory leak -> it's just Oracle not releasing it's
shared
> > memory segments.
> >
> > Check:
> >
> > ipcs -ma
> >
> > and then
> >
> > ipcrm <shmid> to remove unwanted shared memory segments. Of course
beware
> > from removing other processes segments.
> >
> > Tanel.
> >
> > ----- Original Message -----
> > From: "Adolfo" <air_at_tid.es>
> > Newsgroups: comp.databases.oracle.server
> > Sent: Tuesday, September 23, 2003 1:20 PM
> > Subject: Memory leak
> >
> > > Hi,
> > >
> > > We are using oracle server 8.1.7.3 and solaris 2.6. We have put the
> > > latest patchset for solaris 2.6 but when we run a heavy query the
system
> > > memory go down quickly. Our SGA is 300Mb. When we shutdown our
instance
> > > the system not recovery the memory and we have to reset the machine.
> > >
> > > Any ideas?
> > >
> > > Regards,
> > >
>
Received on Tue Sep 23 2003 - 11:18:25 CDT