mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Mike Galbraith <bitbucket@online.de>
To: Michael Wang <wangyun@linux.vnet.ibm.com>
Cc: linux-kernel@vger.kernel.org, mingo@redhat.com,
	peterz@infradead.org, mingo@kernel.org, a.p.zijlstra@chello.nl
Subject: Re: [RFC PATCH 0/2] sched: simplify the select_task_rq_fair()
Date: Thu, 24 Jan 2013 08:47:03 +0100	[thread overview]
Message-ID: <1359013623.5833.52.camel@marge.simpson.net> (raw)
In-Reply-To: <5100DF95.3020909@linux.vnet.ibm.com>

On Thu, 2013-01-24 at 15:15 +0800, Michael Wang wrote: 
> On 01/24/2013 02:51 PM, Mike Galbraith wrote:
> > On Thu, 2013-01-24 at 14:01 +0800, Michael Wang wrote:
> > 
> >> I've enabled WAKE flag on my box like you did, but still can't see
> >> regression, and I've just tested on a power server with 64 cpu, also
> >> failed to reproduce the issue (not compared with virgin yet, but can't
> >> see collapse).
> > 
> > I'm not surprised.  I'm seeing enough inconsistent crap to come to the
> > conclusion that stock scheduler knobs flat can't be used on a largish
> > box, they're just too preempt-happy, leading to weird crap.
> > 
> > My 2 missing nodes came back, and the very same kernel that highly
> > repeatably collapsed with 2 nodes does not with 4 nodes, and 2 nodes
> > does not collapse with only preemption knob tweaking, and that's
> > bullshit.  Virgin shows instability in the mid-range, make a tiny tweak
> > that should have little if any effect there, and that instability
> > vanishes entirely.  Test runs are not consistent enough boot to boot etc
> > etc.  Either stock knobs suck on NUMA boxen, or this box is possessed.
> 
> Mike, I wonder the reason why change back to the old way make collapse
> away may not because there are logical error in new balance path, it's
> just changed the cost of select_task_rq(), whatever it's more or less,
> it's accidentally achieve the same effect as you tweak the knob, so
> that's the reason why it looks like old is better than new.

That's what I'm saying, it's a useless crap side-effect of a preempt
happy kernel.  Results with these knobs are just not stable.  Results go
wildly unstable with 2 nodes vs 4 in this box, but can be stabilized in
all with preemption knob adjustment.. or phase of moon might make them
appear stable.. or not.

-Mike


  reply	other threads:[~2013-01-24  7:47 UTC|newest]

Thread overview: 57+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <1356588535-23251-1-git-send-email-wangyun@linux.vnet.ibm.com>
2013-01-09  9:28 ` Michael Wang
2013-01-12  8:01   ` Mike Galbraith
2013-01-12 10:19     ` Mike Galbraith
2013-01-14  9:21       ` Mike Galbraith
2013-01-15  3:10         ` Michael Wang
2013-01-15  4:52           ` Mike Galbraith
2013-01-15  8:26             ` Michael Wang
2013-01-17  5:55         ` Michael Wang
2013-01-20  4:09           ` Mike Galbraith
2013-01-21  2:50             ` Michael Wang
2013-01-21  4:38               ` Mike Galbraith
2013-01-21  5:07                 ` Michael Wang
2013-01-21  6:42                   ` Mike Galbraith
2013-01-21  7:09                     ` Mike Galbraith
2013-01-21  7:45                       ` Michael Wang
2013-01-21  9:09                         ` Mike Galbraith
2013-01-21  9:22                           ` Michael Wang
2013-01-21  9:44                             ` Mike Galbraith
2013-01-21 10:30                               ` Mike Galbraith
2013-01-22  3:43                               ` Michael Wang
2013-01-22  8:03                                 ` Mike Galbraith
2013-01-22  8:56                                   ` Michael Wang
2013-01-22 11:34                                     ` Mike Galbraith
2013-01-23  3:01                                       ` Michael Wang
2013-01-23  5:02                                         ` Mike Galbraith
2013-01-22 14:41                                     ` Mike Galbraith
2013-01-23  2:44                                       ` Michael Wang
2013-01-23  4:31                                         ` Mike Galbraith
2013-01-23  5:09                                           ` Michael Wang
2013-01-23  6:28                                             ` Mike Galbraith
2013-01-23  7:10                                               ` Michael Wang
2013-01-23  8:20                                                 ` Mike Galbraith
2013-01-23  8:30                                                   ` Michael Wang
2013-01-23  8:49                                                     ` Mike Galbraith
2013-01-23  9:00                                                       ` Michael Wang
2013-01-23  9:18                                                         ` Mike Galbraith
2013-01-23  9:26                                                           ` Michael Wang
2013-01-23  9:37                                                             ` Mike Galbraith
2013-01-23  9:32                                                           ` Mike Galbraith
2013-01-24  6:01                                                             ` Michael Wang
2013-01-24  6:51                                                               ` Mike Galbraith
2013-01-24  7:15                                                                 ` Michael Wang
2013-01-24  7:47                                                                   ` Mike Galbraith [this message]
2013-01-24  8:14                                                                     ` Michael Wang
2013-01-24  9:07                                                                       ` Mike Galbraith
2013-01-24  9:26                                                                         ` Michael Wang
2013-01-24 10:34                                                                           ` Mike Galbraith
2013-01-25  2:14                                                                             ` Michael Wang
2013-01-24  7:00                                                               ` Michael Wang
2013-01-21  7:34                     ` Michael Wang
2013-01-21  8:26                       ` Mike Galbraith
2013-01-21  8:46                         ` Michael Wang
2013-01-21  9:11                           ` Mike Galbraith
2013-01-15  2:46     ` Michael Wang
2013-01-11  8:15 Michael Wang
2013-01-11 10:13 ` Nikunj A Dadhania
2013-01-15  2:20   ` Michael Wang

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1359013623.5833.52.camel@marge.simpson.net \
    --to=bitbucket@online.de \
    --cc=a.p.zijlstra@chello.nl \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@kernel.org \
    --cc=mingo@redhat.com \
    --cc=peterz@infradead.org \
    --cc=wangyun@linux.vnet.ibm.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

Powered by JetHome