Re: [PATCH 11/11] xfs: skip flushing log items during push

["Darrick J. Wong" <djwong@kernel.org>]

On Wed, Jul 03, 2024 at 09:29:47AM +1000, Dave Chinner wrote:
> On Tue, Jul 02, 2024 at 11:51:20AM -0700, Darrick J. Wong wrote:
> > On Fri, Jun 21, 2024 at 10:46:45AM -0700, Darrick J. Wong wrote:
> > > On Fri, Jun 21, 2024 at 07:48:08AM +0200, Christoph Hellwig wrote:
> > > > On Thu, Jun 20, 2024 at 12:51:42PM -0700, Darrick J. Wong wrote:
> > > > > > Further with no backoff we don't need to gather huge delwri lists to
> > > > > > mitigate the impact of backoffs, so we can submit IO more frequently
> > > > > > and reduce the time log items spend in flushing state by breaking
> > > > > > out of the item push loop once we've gathered enough IO to batch
> > > > > > submission effectively.
> > > > > 
> > > > > Is that what the new count > 1000 branch does?
> > > > 
> > > > That's my interpreation anyway.  I'll let Dave chime in if he disagrees.
> 
> Yes, that's correct. I didn't finish this patch - I never wrote the
> comments in the code to explain this because I don't bother doing
> that until I've validated the heuristic and know it mostly works
> as desired. I simply hadn't closed the loop.
> 
> Please add comments to the code to explain what the magic "1000"
> is...

Something along the lines of

	/*
	 * Submit IO more frequently and reduce the time log items spend
	 * in flushing state by breaking out of the item push loop once
	 * we've gathered enough IO to batch submission effectively.
	 */
	if (lip->li_lsn != lsn && count > 1000)
		break;

Maybe?

> > > <nod> I'll await a response on this...
> > 
> > <shrug> No response after 11 days, I'll not hold this up further over a
> > minor point.
> 
> I've been on PTO for the last couple of weeks, and I'm still
> catching up on email. You could have just pinged me on #xfs asking
> if I'd seen this, just like jlayton did about the mgtime stuff last
> week. I answered even though I was on PTO. You always used to do
> this when you wanted an answer to a question - I'm curious as to why
> have you stopped using #xfs to ask questions about code, bugs and
> patch reviews?

I think you told me you had some PTO coming up the month after LSF so I
did not choose to bother you during your time off with something that
didn't seem all /that/ urgent.  At worst, we merge it, try to tweak it,
and either make it better or revert it.

--D

> -Dave.
> -- 
> Dave Chinner
> david@fromorbit.com
>