Re: [PATCH] Give kjournald a IOPRIO_CLASS_RT io priority

[Arjan van de Ven <arjan@infradead.org>]

On Thu, 2 Oct 2008 09:46:59 -0400
Theodore Tso <tytso@mit.edu> wrote:

> On Thu, Oct 02, 2008 at 06:16:29AM -0700, Arjan van de Ven wrote:
> > On Thu, 2 Oct 2008 08:04:44 -0400
> > Theodore Tso <tytso@mit.edu> wrote:
> > 
> > > 
> > > > But sometimes that doesn't work (including the place where I
> > > > knowingly broke it).  If we can find and fix the offending
> > > > piece of jbd logic (a big if) then all is peachy.
> > > 
> > > Do we have workloads that can easily demonstrate this problem?
> > > If so, we can add some tracing code which will allow us to see
> > > which theory is correct, and what is actually happening.
> > 
> > I can very easily reproduce it; my mail client (claws) stalls due to
> > this several seconds at least once every 5 to 10 minutes...
> > (usually when I'm typing an email.. grumble)
> 
> Are you running with noatime or relatime?

yes; I have

mount -o remount,relatime / &
for i in `pidof kjournald` ; do ionice -c1 -p $i ; done

in my /etc/rc.local 


> One quick thing to try that might show what is going on is to run
> debugfs on the device where your mail directory is located, while
> claws is running, and then use the debugfs command "logdump -a" to
> dump out the contents of the journal.  You can then use ncheck and
> icheck to take the FS block numbers from logdump and translate them to
> inode numbers, and then from inode numbers to pathnames.  That might
> give us some insight as to what is going on.

I'll try (sadly I'm in a training session all day today so I might not
get to it quickly)
 
> I can whip up a patch which adds some markers which we could use to
> find out more information about what is happening.

interesting testcase of the markers concept.


-- 
Arjan van de Ven 	Intel Open Source Technology Centre
For development, discussion and tips for power savings, 
visit http://www.lesswatts.org