Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!munnari.OZ.AU!news.ecn.uoknor.edu!feed1.news.erols.com!cpk-news-hub1.bbnplanet.com!news.bbnplanet.com!newsxfer3.itd.umich.edu!news1.best.com!nntp1.ba.best.com!not-for-mail
From: dillon@flea.best.net (Matt Dillon)
Newsgroups: comp.unix.bsd.freebsd.misc,comp.sys.sgi.misc
Subject: Re: no such thing as a "general user community"
Date: 27 Mar 1997 13:01:26 -0800
Organization: BEST Internet Communications, Inc.
Lines: 33
Message-ID: <5hen76$77l@flea.best.net>
References: <331BB7DD.28EC@net5.net> <5h9vft$8eo@fido.asd.sgi.com> <5han4u$fnf@flea.best.net> <gergE7ouGw.HK8@netcom.com>
NNTP-Posting-Host: flea.best.net
Xref: euryale.cc.adfa.oz.au comp.unix.bsd.freebsd.misc:37974 comp.sys.sgi.misc:29494
:In article <gergE7ouGw.HK8@netcom.com>, Greg Andrews <gerg@netcom.com> wrote:
:>dillon@flea.best.net (Matt Dillon) writes:
:>>
:>> that we STILL have a once-a-minute cron job running on the two L's which
:>> allocates 130 MBytes of ram, touches it all, then exits.
:>>
:>
:>Out of curiosity, what does that do? Flush pages to disk and then
:>force them back into the free pool?
:>
:> -Greg
You don't wanna know. What happens is that in the space of a few
seconds, the efficiency of the paging system will drop by three orders
of magnitude. Sometimes we observe it paging out at insanely slow
rates... 16K/sec or 32K/sec rather then (bursts of) 4 and 5 MBytes/sec
which is what it should be doing. SGI supposedly fixed that problem,
and indeed we do not see such low paging rates as often, but it hasn't
fixed the multi-magnitude drop in efficiency problem. It's either
paging out the wrong pages entirely, not paging out enough pages,
or getting into weird sequenced lockout conditions in the kernel
whereby some degenerate page demand condition is preventing the pager
from paging out pages. I don't know which. I suspect it is related
to the large number of open vnodes in the systems. There are so
many hacks in the IRIX swapper/pager code that they are probably tripping
over their own feet.
The only way to pop the paging system back into normal operation is to
literally force the free memory to 0 for several seconds. This occurs
on all of our heavily loaded SGI's. The only way to do this is to
allocate and touch a whole lot of memory, then free it.
-Matt