*BSD News Article 20058


Return to BSD News archive

Newsgroups: comp.os.386bsd.questions
Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!elroy.jpl.nasa.gov!usc!cs.utexas.edu!geraldo.cc.utexas.edu!portal.austin.ibm.com!willcox!obiwan!bob
From: bob@obiwan.uucp (Bob Willcox)
Subject: Re: swap space, X server for syscons-0.2
Organization: Bob's Place, Austin TX
References: <CCBL2A.IyH@latcs1.lat.oz.au>
Message-ID: <CCDC3B.M03@obiwan.uucp>
Date: Thu, 26 Aug 1993 13:28:08 GMT
Lines: 77

In article <CCBL2A.IyH@latcs1.lat.oz.au>,
M.C. Wong <wongm@latcs1.lat.oz.au> wrote:
>Hi,
>  In a previous posting, I get some helpful replies by some of those kind souls.
>And I got swapinfo to check on the use of my swap partition. And as it reveals,
>5 MB is definitely too small. But there is one thing I do not quite understand
>the way 386bsd does virtual memory  management. I have tried running some
>applications (like xfig, ghostview, etc) and did swapinfo before, during and 
>after those app. execution. Initially, my system is not fully loaded to need
>any swapping, when those app are loaded, swapinfo reveals about 80% - 100% of
>swap space used (varied depending on which app.). But when I quit those new
>app, and did swapinfo again, it reveals that the the swap space in use it still
>the same as last revealed while the app was running. And, of course I try 
>activate every program on the screen (from X) to see if any of those swapped-out
>pages are swapped-in again, by running swapinfo again, but it didn't show any
>change on the swap space used. I wonder if this is the behaviour (ie the pages
>that swapped out are not my on-scree programs but rather system daemons etc?)
>expected ? Or should I see a 0% capacity , ie back to original state before I
>ran those extra app. (xfig, ghostview etc) ?

Try running "ps aux" before and after you run the applicationss and you
should notice in the after ps that there is a significant difference
between the VSZ & RSS columns.  I believe that this reflects the pages
that are paged out.

>
>  Secondly, someone points me to fetch an updated of X server for XF86_SVGA
>that has bug fix for line drawing function. I wonder does there exis a version
>for syscons-0.2 as well ? I am running syscons-0.2 and feel that its performance
>is better than pccons and codrv, I wonder will syscons makes its way into 
>XFree-2.0 ?

I don't know anything about this.

>
>  One last question, I remember reading some general guidelines/formula for
>allocate swap space from the book "Unix Administration Handbook", but can't
>remember exactly what it is, and I don't have access to that book anymore, I
>wonder if someone can let me know ? 

There was a rather lengthy discussion on this awhile back.  I don't
remember what or if a conclusion was reached.  Since it seems that
386bsd does not deal gracelfully with running out of swap space, I
have simply made my total swap space pretty big (I hope :-).  Here
is output from "swapinfo -k" on my system (w/20MB of real):

Device      Kilobytes       Used  Available   Capacity
/dev/wd0b       20588       3968      16620      19%
/dev/wd1b       20668       4096      16572      20%
/dev/as0b       24736       2576      22160      10%
/dev/as1b       24620       1776      22844       7%
Total           90612      12416      78196      14%

>Also, I plan to disklabel my new disk
>in such a way that I am leaving some "hole" between parition b and other 
>paritions to leave some room for grow for the new additional swap partition,
>is that feasible ? Has anyone done it before ? 

Hmm, why not just disklabel a "b" partition the size you want it
to be for the added swap (b is, at least, the convention for the
swap partition).  Don't forget to update your configruation file
to specify the additional swap disk, rebuild your kernel, and add
the swap partition to your /etc/fstab file.

>How about if I use the extra
>hole to augment an existing file system ?

If you mean ``can I increase the size of an existing file system
w/o rebuilding it'' (back it up, re-disklabel to new size, newfs,
restore) the answer is no.  You could create another file system
in the hole though.

I hope some of this helps,
-- 
Bob Willcox                ...!{rutgers|ames}!cs.utexas.edu!uudell!obiwan!bob
Phone: 512 258-4224 (home)
       512 838-3914 (work)