NetBSD Problem Report #46952

From dholland@macaran.localdomain  Fri Sep 14 02:47:38 2012
Return-Path: <dholland@macaran.localdomain>
Received: from mail.netbsd.org (mail.netbsd.org [149.20.53.66])
	by www.NetBSD.org (Postfix) with ESMTP id A926C63B9BC
	for <gnats-bugs@gnats.NetBSD.org>; Fri, 14 Sep 2012 02:47:38 +0000 (UTC)
Message-Id: <20120914024744.28EEE6E230@macaran.localdomain>
Date: Thu, 13 Sep 2012 22:47:43 -0400 (EDT)
From: dholland@eecs.harvard.edu
Reply-To: dholland@eecs.harvard.edu
To: gnats-bugs@gnats.NetBSD.org
Subject: VM page allocation is borked
X-Send-Pr-Version: 3.95

>Number:         46952
>Category:       kern
>Synopsis:       VM page allocation is borked
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri Sep 14 02:50:00 +0000 2012
>Originator:     David A. Holland
>Release:        NetBSD 6.99.11 (20120906)
>Organization:
>Environment:
System: NetBSD macaran 6.99.11 NetBSD 6.99.11 (MACARAN) #13: Thu Sep 6 20:22:11 EDT 2012 root@macaran:/usr/src/sys/arch/amd64/compile/MACARAN amd64
Architecture: x86_64
Machine: amd64
>Description:

I am running three concurrent copies of a simulation process, each
consuming about 1.5G, and even though I have 6G of RAM it is swapping
to death.

On tls's suggestion I tried raising vm.anonmax from 80 to 95; this
helped a lot. Lowering vm.execmin and vm.filemin to 3 helped a bit
more. However, even with these measures the system continued swapping;
just not as hard.

This workload should be fitting in memory. (And while I have an X
server and a browser running, they should be, and AFAICT are, mostly
paged out; in any event I'm 2G into swap.)

This may be a regression, as the reason I chose to run this thing in
batches of three is that prior experience, some a few months ago with
6.99.3 and more a bit further back using 5.99.59, was that three
copies would fit in memory without swapping but four wouldn't.

There seem to be two problems here: one that a workload using 75% of
installed RAM swaps, and one that the VM tuning is making it quite a
bit worse.

>How-To-Repeat:

Run a few processes with large working sets.

>Fix:

fiik.

NetBSD Home
NetBSD PR Database Search

(Contact us) $NetBSD: query-full-pr,v 1.39 2013/11/01 18:47:49 spz Exp $
$NetBSD: gnats_config.sh,v 1.8 2006/05/07 09:23:38 tsutsui Exp $
Copyright © 1994-2007 The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.