[Ocfs-users] OCFS Database too slow

Wed Jul 21 09:33:42 CDT 2004

Hi Wim...

I have monitor by using VMSTAT, and iowaits is giving 3 digits, i think it's
quite high. I have tried to reduce to 13 LUNS in my TEST environment with
the same HW configuration as PROD environment, but not giving big changes.
Now is there any other way to avoid using SECUREPATH?

Rgds/Jeram

-----Original Message-----
From: Wim Coekaerts [mailto:wim.coekaerts at oracle.com]
Sent: Wednesday, July 21, 2004 8:24 AM
To: Jeram
Cc: Wim Coekaerts; ocfs-users at oss.oracle.com
Subject: Re: [Ocfs-users] OCFS Database too slow

i thik the luns can become a bottleneck for one.
i'd also try to not use securepath for 1 time and see how that changes
performance and then go back if it's the same etc

people should run vmstat for a while without anything running , just
have the filesystems monted and see waht the amount of iowait on the box
is 

On Wed, Jul 21, 2004 at 08:12:43AM +0700, Jeram wrote:
> Hi Wim...
> 
> This discussion is quite interesting, I am having some performance issue,
> but till now still under observation, This is my environment details :
> 1. We are using RHAS 2.1 with kernel 2.4.9-e.27 Enterprise
> 2. OCFS version : 2.4.9-e-enterprise-1.0.9-6
> 3. Oracle RDBMS : 9.2.0.4 RAC with 5 Nodes
> 4. Storage = EVA 6000 with 8 TB SIZE
> 5. We have 1 DiskGroup and 51 LUNs configured in EVA6000.
> 
> You are talking about securepath, we are using securepath right now, is
> there any consideration or suggestion if we are using securepath ? I have
> tried to upgrade to OCFS 1.0.12, but i dont see any performance
improvement.
> 
> Thanks
> Jeram
> -----Original Message-----
> From: Wim Coekaerts [mailto:wim.coekaerts at oracle.com]
> Sent: Saturday, July 17, 2004 12:42 AM
> To: Varghese Abraham
> Cc: ocfs-users at oss.oracle.com
> Subject: Re: [Ocfs-users] OCFS Database too slow
> 
> 
> are you using securepath ?
> 
> On Fri, Jul 16, 2004 at 10:29:47AM -0700, Varghese Abraham wrote:
> > 
> > Hi Wim,
> >  I appreciate your reply and I have been following this mailing list
very
> closely for the last 3-4 months.
> > 
> > I am not at all trying to compare between ext3 and ocfs.
> > 
> > The current installation that we are having is dead slow..
> > 
> > 
> > We had a previous rac installation (TEST1 ) with OCFS ( kernel e27 ,
ocfs
> 1.0.9 , 9.2.0.3 database).
> > On this old system with OCFS, our performance was beautiful.
> > 
> > 
> > I am not able to get even a foot close to the previous systems  ( TEST
1)
> performance.
> > 
> > Neither am I blaming OCFS.
> >  After having seen good performance with TEST1 system, I am unable to
make
> out what has gone wrong with the new system although we are on the latest
> for everything with the new system.
> > 
> > I am sure it is some small issue and that is what I need your help in
> trying to debug.
> > 
> > R'gds
> > Varghese Abraham.
> > 
> > 
> > 
> > 
> > 
> > -----Original Message-----
> > From: wim.coekaerts at oracle.com [mailto:wim.coekaerts at oracle.com]
> > Sent: Friday, July 16, 2004 10:13 AM
> > To: Varghese Abraham
> > Cc: ocfs-users at oss.oracle.com
> > Subject: Re: [Ocfs-users] OCFS Database too slow
> > 
> > 
> > yes you are probably comparing local filesystem versus raw or cfs which
> > is not cached in the OS only in oracle.
> > 
> > this is a story that comes back on this list every few weeks or months
> > and basically happens because there is a huge misunderstanding on what
> > the OS filesystem cache provides single node etc etc, it takes time to
> > explain and I think wehave in the past. could you go through the
> > archives and look for earlier discussions on this ? 
> > 
> > there is nothing slow, what youa re doing is having a database run that
> > has a lot of stuff cached in the OS itself on local fielsystem and then
> > when you do that query on cfs or raw it has to go to disk, the rigth
> > thing to do (should do) is give all that memory to oracle. then you will
> > see the difference
> > 
> > again, it's not abug, don't think it's slow, just do the right thing.
> > and look in archives for a pointer to the discussions
> > 
> > Wim
> > 
> > On Fri, Jul 16, 2004 at 09:39:33AM -0700, Varghese Abraham wrote:
> > > 
> > > Hi All,
> > >  we are using Red Hat 2.1 Kernel e38 along with MSA 1000.
> > > ocfs version being used is
> > > 
> > > $ rpm -qa | grep ocfs
> > > ocfs-tools-1.0.10-1
> > > ocfs-2.4.9-e-enterprise-1.0.12-1
> > > ocfs-support-1.0.10-1
> > >  
> > > 
> > > Database Version is 9.2.0.5
> > > 
> > > 
> > > However we find that the performance of the database on OCFS is too
> slow. even a select count(1) from all_tables takes like a while to
complete.
> > > 
> > > We initially assumed RAC is the problem area.  ( Because with both
nodes
> enabled we had terrible performance)
> > > 
> > > Hence we stopped one of the RAC Databases to test out the performance.
> > > However even with one node the performance is real bad.
> > > 
> > >  
> > > 
> > > Any pointers on how to debug this problem.
> > > 
> > > 
> > > R'gds
> > > Varghese Abraham.
> > > _______________________________________________
> > > Ocfs-users mailing list
> > > Ocfs-users at oss.oracle.com
> > > http://oss.oracle.com/mailman/listinfo/ocfs-users
> > 
> 
> _______________________________________________
> Ocfs-users mailing list
> Ocfs-users at oss.oracle.com
> http://oss.oracle.com/mailman/listinfo/ocfs-users