On Wed, Jan 30, 2013 at 11:07 AM, Jan Kara [off-list ref] wrote:

On Wed 30-01-13 00:26:58, Ted Tso wrote:

quoted

On Wed, Jan 23, 2013 at 08:29:11PM +0100, Jan Kara wrote:

quoted

quoted

old JBD: AIM7 jobs/min 97624.39;  got 78193 jbd wakeups
new JBD: AIM7 jobs/min 85929.43;  got 6306999 jbd wakeups, 6264684 extra wakeups

  Yeah, that's a lot. My guess would be *a lot* of processes are hanging in
start_this_handle() and waiting for transaction commit. Each of them calls
__log_start_commit() and things add up. Thanks for getting these numbers.

Yeah, wow.  That would imply that there are a huge number of processes
that get hung up in start_this_handle(), and they are waking up the
journal before the kjournald has a chance to set T_LOCKED (since then
they would get blocked earlier in start_this_handle).

Given that, I wonder if the following change would actually help or
hurt things.  Eric, would you be willing to ask your perf team to try
testing out these patches?

  Umm, I don't see anything. Forgot to attach them?

Here I catched the two patches:

http://patchwork.ozlabs.org/patch/216768/
http://patchwork.ozlabs.org/patch/216767/

- Sedat -

                                                                Honza
--
Jan Kara [off-list ref]
SUSE Labs, CR