[Ocfs2-users] Anyone have an idea how to find file i/o throughput?

[Andrew Phillips]

Ulf,

  Have you considered using systemtap? There is a recipe here that could
be used to find out whats going on;

http://sourceware.org/systemtap/wiki/WSDeviceMonitor?highlight=%28%
28WarStories%29%29

   I'm not sure how well that would work with ocfs2. Unlike dtrace,
systemtap can be more "uneven" in coverage. Its also something that
requires a bit of fiddling (installing debuginfo packages).

   The recipe above traps vfs_read and vfs_write so should work as a 
first stab at identifying the process id thats causing the I/O. 

   I'd also advise some thought if its to be used on a production
environment. Having said that, I've used it on a production oracle RAC
database server and found it very valuable.

   I don't recall you mentioning the distribution, but RH, CentOS, and 
oracle's version of CentOS should all work. 

   As always, read the instructions on the label, etc...

     Andy
   

On Mon, 2008-02-18 at 23:14 -0800, Ulf Zimmermann wrote:
> Forgot to mention, this remote server is just Oracle. It has one standby
> database and one local database, the local one is suppose to be idle,
> i.e. nothing connecting to it, besides once in a while for available
> check.
> 
> While the primary database of the standby was down, I saw less disk read
> access, but every 5 minutes for about 60 seconds I would see
> 50-60MB/sec. After the primary came back up, read access is as high as
> 160MB/sec.
> 
> We are only seeing it on this single node of the remote standby. The
> local standby (on EXT3) is not doing the same thing.
> 
> > -----Original Message-----
> > From: Sunil Mushran [mailto:sunil.mushran at oracle.com]
> > Sent: Monday, February 18, 2008 19:28
> > To: Ulf Zimmermann
> > Cc: ocfs2-users at oss.oracle.com
> > Subject: Re: [Ocfs2-users] Anyone have an idea how to find file i/o
> > throughput?
> > 
> > If a userspace process is behind the io surge, then strace should
> help.
> > But determining the process may require a bit of trial and error.
> > 
> > Ulf Zimmermann wrote:
> > > We got a remote Oracle 10g R2 standby running on OCFS2. Initial when
> we
> > > started the standby, read I/O was < 5MB/sec on average. Since then
> it
> > > has grown to over 40MB/sec (longer average, it peaks much higher).
> Here
> > > is a graph showing this:
> > >
> > > http://www.alameda.net/~ulf/dbphx01.png
> > >
> > > We also have a local standby running (on EXT3) which is not showing
> the
> > > same symptom. I am trying to find where all these reads are
> happening.
> > > Anyone have an idea how to figure that out on Linux?
> > >
> > > Ulf.
> > >
> > >
> > > _______________________________________________
> > > Ocfs2-users mailing list
> > > Ocfs2-users at oss.oracle.com
> > > http://oss.oracle.com/mailman/listinfo/ocfs2-users
> > >
> 
> 
> _______________________________________________
> Ocfs2-users mailing list
> Ocfs2-users at oss.oracle.com
> http://oss.oracle.com/mailman/listinfo/ocfs2-users

________________________________________________________________________
In order to protect our email recipients, Betfair Group use SkyScan from 
MessageLabs to scan all Incoming and Outgoing mail for viruses.

________________________________________________________________________