SUMMARY:swap slice vs swapfile

From: Mangan, Paul IT-IAU <Paul.Mangan_at_Heidelberg.com>
Date: Wed Nov 05 2003 - 10:23:47 EST
Initial request :" ...for future reference can someone inform me of the pros
and cons of a swap slice vs a swapfile on a mounted slice"

First of all "Thank You" to all who have given me such great advice, opinions
and facts:
Rich Kulawiec
Michael T Pins
Brett Lymn
Russell C Page
Sandor Orban
Warren Powell
Eugene Schmidt
JV
Anthony A. D. Talltree
Kevin Buterbaugh
Jose Luis Martinez
Tim Chipman
Craig Touchton
Orville M. Lewis
Reggie Beavers
WC Jones
Steven Haywood

Suggested Reading:
man pages on dumpadm
Unix Introduction Manuals RE:filesystems and virtual memory

Suggested commands:
dumpadm
swap -l
swap -s
cat /etc/vfstab

The consensus of opinions is that a dedicated swap slice is faster than a
swapfile on a filesystem slice. This agrees with everything I have learned in
SUN Solaris training classes.

The most negative affect that was discussed is that there is a belief that
during a kernel dump there would be no write to the swapfile and it is
believed that SUN recommends that you have a swap partition to catch kernel
dumps. This can be interpreted to be backed up by SUN if you read the dumpadm
man pages.

AND NOW FOR SOME FACTS:
While receiving responses as an experiment I built one V240 WITHOUT a swap
slice and WITHOUT a swapfile. I used slice 0 as root, slice 4 as a data area
that could contain a swapfile and slice 5 as a future data area. I have 512 MB
of ram and a 36gig hard disk divided equally amongst the slices. I installed
Solaris 9, August 2003 release using the RSC feature via Ethernet and a local
DVD on the V240. When the system completed the installation it rebooted
without a problem which left me wondering how it could do that without a swap
area. Solaris 1 or SunOS 4.X used to have a problem if you tried that. Upon
investigation I used the command swap -l and discovered as expected there was
no swap device configured. Then I used the command swap -s and found I had
about 400meg of swap space available and 23meg was being used. Further
investigation indicated that the system had automatically set up /tmp and
/var/run as the swap areas each with 375 meg of space that it could use. The
vfstab indicated that a swap area had been set up using tmpfs on /tmp.

So the basic info is that it can be built without a swap slice or a swapfile
and the installation procedure will take over and create a bootable system
using the above areas. The system functions reasonably and whether it will
create a dumpfile of value for the kernel, etc. is not known.

Thanks to everyone for the input which I strongly value. I hope my experiment
returns some useful information to each of you.

Paul Mangan
mangan@sprynet.com
paul.mangan@heidelberg.com
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
Received on Wed Nov 5 10:28:31 2003

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:43:23 EST