Re: AWE implementation on Windows 2000 - Potential solution for intermittent TNS-12560
- From: Guillermo Alan Bort <cicciuxdba@xxxxxxxxx>
- To: "Crisler, Jon" <Jon.Crisler@xxxxxxx>
- Date: Tue, 12 May 2009 22:48:39 -0300
This is a Siebel server, and for what I've been seeing, it doesn't even use
a connection pool... There is no apps dba, so I'm still investigating the
issue...
MTS is out of the question. I'd use it if it was a somewhat stable system
and we had a similar environment to test. But this is a 9i RAC on windows
2000... I can think of so many bugs that adding MTS would be inviting
trouble.
For now, we will wait and see what Oracle Support comes up with. The best
idea the've come up with so far is to reduce the SGA components so there is
more memory for user processes.... I thought of it some time ago, but it
requires some fine tuning work, which takes time... and I'm already in the
process of doing... so nothing interesting so far.
Alan Bort
Oracle Certified Professional
On Tue, May 12, 2009 at 1:02 PM, Crisler, Jon <Jon.Crisler@xxxxxxx> wrote:
> The theory is that shared server connections (MTS) should use less
> memory. However, I have been burned so many times with bugs and
> oddities, in everything from Oracle 7.3 to 10g, that I rarely use MTS.
> Our standard is not to use it, and if we do use MTS we run a stress test
> that is supposed to reveal any issues (and that test has mixed results).
>
> That said, I love the idea of MTS and if the situation warrants it, you
> should try it.
>
> -----Original Message-----
> From: Niall Litchfield [mailto:niall.litchfield@xxxxxxxxx]
> Sent: Tuesday, May 12, 2009 11:53 AM
> To: cicciuxdba@xxxxxxxxx; Crisler, Jon; oracle-l-freelists
> Subject: Re: AWE implementation on Windows 2000 - Potential solution for
> intermittent TNS-12560
>
> Hi do your users use dedicated or shared server connections? You may
> find that using shared servers reduces the memory pressure on the
> oracle.exe process.
>
> On 5/11/09, Guillermo Alan Bort <cicciuxdba@xxxxxxxxx> wrote:
> > I know the upgrade options I can take, the problem is the upgrade plan
> is
> > scheduled for the end of the year, and If I have to go through the
> next
> > eight monthgs without sleeping because of these database, I'll murder
> > someone. So I need a somewhat short term solution. I keep reading
> about
> > this, but cannot find anything. I don't want to enable the 3G switch
> because
> > it could cause more instance evictions... At this point, I don't
> really know
> > what to do.
> >
> > Thanks anyway.
> > Alan Bort
> > Oracle Certified Professional
> >
> >
> > On Mon, May 11, 2009 at 4:46 PM, Crisler, Jon <Jon.Crisler@xxxxxxx>
> wrote:
> >
> >> Its been a while since I worked on these versions, but perhaps you
> are
> >> running into a limit of 9i 32 bit ? For instance, 32 bit 9i and 10g
> on
> >> RedHat Linux have a practical limit of 1.7gb in the sga, unless you
> resort
> >> to tricks like hugemem, hugepages or hugetlb in the kernel, then
> relink
> >> Oracle 9i (no need to relink 10g).
> >>
> >>
> >>
> >> Assuming that Siebel supports 10g, this would be a perfect candidate
> for
> >> Oracle 10g RAC on Linux.
> >>
> >>
> >> ------------------------------
> >>
> >> *From:* oracle-l-bounce@xxxxxxxxxxxxx [mailto:
> >> oracle-l-bounce@xxxxxxxxxxxxx] *On Behalf Of *Guillermo Alan Bort
> >> *Sent:* Friday, May 08, 2009 6:25 PM
> >> *To:* oracle-l-freelists
> >> *Subject:* AWE implementation on Windows 2000 - Potential solution
> for
> >> intermittent TNS-12560
> >>
> >>
> >>
> >> Listers,
> >>
> >> I have a two node 9i RAC on Windows 2000 (SP4) Advanced Server.
> This
> >> database serves a Siebel software running on a different machine. I
> don't
> >> know why these are windows machines, and I have no prior experience
> with
> >> RAC
> >> on windows. I know some of you might, but I'm not so sure about 9i.
> >>
> >> As this is 9i (9.2.0.8.0) ASM is out of the question. This
> particular
> >> setup is based on OCFS (OCSF1, of course) with the latest patchset
> applied
> >> to the srvm.
> >>
> >> Now, the problem started some time ago with frequent instance
> evictions
> >> (actually the version back then was 9.2.0.5.0) becoming more frequent
> >> after
> >> an anti-virus upgrade. I have requested to add oracle osd9i and
> >> oracle_home
> >> to the exception list of the AV, but with no luck so far (they have
> not
> >> done
> >> so yet). I applied 9.2.0.8.0 both to CRM and OH in order to be in a
> >> supported level, and opened an S.R. with Oracle.
> >>
> >> Instance evictions have actually dimished considerably (if not
> stopped)
> >> but a new problem arose. Since the upgrade we have been getting
> several
> >> TNS-12560 Protocol Adapter Error from one of the nodes (the second
> one)
> >> which is the one most users use to connect. After opening yet another
> SR
> >> with Oracle, they suggested that I should increase the buffer cache.
> Now,
> >> this is windows 2000, with 6GB (and obviously PAE wnabled at the OS
> level)
> >> but I've been unable to extend the db_cache beyond the 2G limit.(I
> did
> >> indeed use indirect_data_buffers and the legacy parameter
> >> db_block_buffers).
> >> Even so, I'm not able to extend the SGA beyond 2G.
> >>
> >> I've researched metalink and I'm most certain I'm missing
> something. so
> >> I was wondering what you take on this is, and if you have any ideas.
> >>
> >> On the other hand, I will get a SP report to make sure we have no
> wild
> >> queries and try to install OSW to monitor os performance.
> >>
> >> Thanks in advance.
> >> Alan Bort
> >> Oracle Certified Professional (just in 10g)
> >>
> >
>
> --
> Sent from Google Mail for mobile | mobile.google.com
>
> Niall Litchfield
> Oracle DBA
> http://www.orawin.info
>
Other related posts:
