Summary: Unified / and /usr: good or bad idea?

From: Ralf W. Grosse-Kunstleve <rwgk_at_laplace.csb.yale.edu>
Date: Thu, 26 Feb 1998 14:58:59 -0500 (EST)

Thank you for the responses to my question about a unified / and /usr
partition!
I append all messages I got.
I decided to go back to two different file systems for / and /usr.
The reason is that I want to make sure that the upgrade and/or patch
tools work.
I have good experiences working with a common / and /usr filesystem
on HP and SGI platforms.
Even if the root filesystem is separated, users can still fill it up
by writing to /tmp. If the root filesystem is small, it can happen
very quickly.
The solution is of course to use the newfs -m minfree option.
I guess this would work just as well on a unified filesystem.
Ralf





>From risler_at_genetique.uvsq.fr Wed Feb 25 13:17:18 EST 1998
Sender: risler_at_uvsq.fr
Date: Wed, 25 Feb 1998 19:17:00 +0100
From: Jean-Loup Risler <risler_at_genetique.uvsq.fr>
To: "Ralf W. Grosse-Kunstleve" <rwgk_at_laplace.csb.yale.edu>
Subject: Re: Unified / and /usr: good or bad idea?

Hi,

Following your question: I think it IS a bad idea....

One day or another, you will find your /usr full, not speaking of
/var and /tmp ..... If this happens, you will end up with a 110%
full root file system and you will be blocked solid.

By far, the best method (not only for DU) is to have separate file
systems for /, /usr, /var and /tmp.

Hope it helps.....

Jean-Loup
-- 
---------------------------------------------------------------------
| Jean-Loup Risler               |                                  |
| Universite de Versailles       | Tel: 33 (0)1 39 25 45 54         |
| Lab. Genome et Informatique    | Fax: 33 (0)1 39 25 45 69         |
| Batiment Buffon                |                                  |
| 45 Avenue des Etats-Unis       | email: risler_at_genetique.uvsq.fr  |
| 78035  Versailles Cedex France |                                  |
---------------------------------------------------------------------
Les cerveaux sont malencontreusement pourvus d'estomac (J. Perrin)
>From alan_at_nabeth.cxo.dec.com Wed Feb 25 13:28:56 EST 1998
From: alan_at_nabeth.cxo.dec.com
To: "Ralf W. Grosse-Kunstleve" <rwgk_at_laplace.csb.yale.edu>
Subject: Re: Unified / and /usr: good or bad idea? 
Date: Wed, 25 Feb 98 11:28:38 -0700
	I think the main reason for the split is tradition; it has
	always been done that way.  The split of /usr and /var is
	necessary in an environment where multiple disk-less systems
	share the same /usr.  The parts in /var are read-write and
	very system dependent.  The parts in a split /usr can be
	safely shared.  The same is true for the root since some
	configurations files are in /etc and the kernels may be
	different.  For a completely standalone system, neither
	of these is likely to be an issue.
	One of advantage of splitting the file systems is that
	a disk of the appropriate size and performance characteristics
	can be selected for the expected load.  Ignoring the C2
	security stuff (which I don't know enough about), once
	/tmp is moved off the root file system, the load to it
	tends to be read-only.  RAID-5 or Mirroring may be very
	appropriate for providing highly available storage.  The
	same is true for /usr when /var is moved off.  /var on
	the other hand will see lots of writes and RAID-5 may not
	be appropriate.
	Also, the root is small enough and a good sized buffer cache
	effective enough that the load to / might be very small since
	the read-mostly files will tend to be cached.
	The root is the set of programs need for doing basic maintenance
	on a system.  Should a disk go bad, there is a small chance that
	the root file system will be affected when it is small (based on
	the size).  By spreading the root out over the whole disk, an
	error anywhere on the disk could affect the important parts
	that you need for running it standalone.
	Finally, it seems unlikely that the engineers responsible for
	laying out the system have ever tested it this way.  Such a
	configuration is likely unsupported and if you have any
	trouble with it, you're on your own.  Good luck.
>From Roger.Picard_at_digital.com Wed Feb 25 13:29:40 EST 1998
From: Roger Picard <Roger.Picard_at_digital.com>
To: "'Ralf W. Grosse-Kunstleve'" <rwgk_at_laplace.csb.yale.edu>
Subject: RE: Unified / and /usr: good or bad idea?
Date: Wed, 25 Feb 1998 13:12:14 -0500
Ralf,
	A very bad idea.   Any run-away process that logs errors will
	fill the file system and choke the system if / is allowed to
fill up.
	you may overwrite critical blocks on your disk.
	if users have no disk quotas then they could copy a large file
	until the system chokes.
Roger
	----------
	From: 	Ralf W. Grosse-Kunstleve[SMTP:rwgk_at_laplace.csb.yale.edu]
	Reply To: 	Ralf W. Grosse-Kunstleve
	Sent: 	Wednesday, February 25, 1998 12:49 PM
	To: 	alpha-osf-managers_at_ornl.gov
	Cc: 	rwgk_at_laplace.csb.yale.edu
	Subject: 	Unified / and /usr: good or bad idea?
	I am running 4.0C.
	I copied the whole system with dump/restore to an external disk
with
	only one ufs partition (and one swap partition). Booting that
	system seemed ok, so I repartitioned the internal drive to also
have
	only the rz*a and rz*b partitions and copied the whole system
back.
	So far everything seems to work just fine. Therefore I was
wondering
	why the default is to have the disk artificially fragmented.
	Is there a bad catch when running DEC Unix with unified / and
/usr
	partitions?
	Ralf
	P.S.: here is what I did (twice) to copy the whole system to
another disk:
	    newfs /dev/rrz29a /dev/rrz29a
	    mount /dev/rz29a /mnt
	    dump 0f - / | ( cd /mnt ; restore rf - )
	    dump 0f - /usr | ( cd /mnt/usr ; restore rf - )
	    cp -p /mnt/genvmunix /mnt/vmunix
	    edit /mnt/etc/fstab: change definition of root and swap
device
	    rm /mnt/sbin/swapdefault
	    shutdown -h now
	    boot dkd500.5.0.1010.0
	    Rebuild kernel (just to be clean)
>From Alan.Davis_at_digital.com Wed Feb 25 13:44:13 EST 1998
From: Alan Davis <Alan.Davis_at_digital.com>
To: "Ralf W. Grosse-Kunstleve" <rwgk_at_laplace.csb.yale.edu>
Subject: RE: Unified / and /usr: good or bad idea?
Date: Wed, 25 Feb 1998 13:11:13 -0500
This message is in MIME format. Since your mail reader does not understand
this format, some or all of this message may not be legible.
------ =_NextPart_000_01BD4218.C5457EDA
Content-Type: text/plain
Ralf,
   Combining / and /usr is not supported on DU. Lot's of possibility for
breakage, not the least of which is that installupdate will not work.
Alan Davis
	-----Original Message-----
	From:	Ralf W. Grosse-Kunstleve
[SMTP:rwgk_at_laplace.csb.yale.edu]
	Sent:	Wednesday, February 25, 1998 12:50 PM
	To:	alpha-osf-managers_at_ornl.gov
	Cc:	rwgk_at_laplace.csb.yale.edu
	Subject:	Unified / and /usr: good or bad idea?
	I am running 4.0C.
	I copied the whole system with dump/restore to an external disk
with
	only one ufs partition (and one swap partition). Booting that
	system seemed ok, so I repartitioned the internal drive to also
have
	only the rz*a and rz*b partitions and copied the whole system
back.
	So far everything seems to work just fine. Therefore I was
wondering
	why the default is to have the disk artificially fragmented.
	Is there a bad catch when running DEC Unix with unified / and
/usr
	partitions?
	Ralf
	P.S.: here is what I did (twice) to copy the whole system to
another disk:
	    newfs /dev/rrz29a /dev/rrz29a
	    mount /dev/rz29a /mnt
	    dump 0f - / | ( cd /mnt ; restore rf - )
	    dump 0f - /usr | ( cd /mnt/usr ; restore rf - )
	    cp -p /mnt/genvmunix /mnt/vmunix
	    edit /mnt/etc/fstab: change definition of root and swap
device
	    rm /mnt/sbin/swapdefault
	    shutdown -h now
	    boot dkd500.5.0.1010.0
	    Rebuild kernel (just to be clean)
------ =_NextPart_000_01BD4218.C5457EDA
Content-Type: application/ms-tnef
Content-Transfer-Encoding: base64
------ =_NextPart_000_01BD4218.C5457EDA--
>From tpb_at_zk3.dec.com Wed Feb 25 13:48:39 EST 1998
To: "Ralf W. Grosse-Kunstleve" <rwgk_at_laplace.csb.yale.edu>
Subject: Re: Unified / and /usr: good or bad idea? 
Date: Wed, 25 Feb 98 13:46:58 -0500
From: "Dr. Tom Blinn, 603-884-0646" <tpb_at_zk3.dec.com>
> I am running 4.0C.
> I copied the whole system with dump/restore to an external disk with
> only one ufs partition (and one swap partition). Booting that
> system seemed ok, so I repartitioned the internal drive to also have
> only the rz*a and rz*b partitions and copied the whole system back.
> So far everything seems to work just fine. Therefore I was wondering
> why the default is to have the disk artificially fragmented.
> Is there a bad catch when running DEC Unix with unified / and /usr
> partitions?
> Ralf
> 
> P.S.: here is what I did (twice) to copy the whole system to another disk:
>     newfs /dev/rrz29a /dev/rrz29a
>     mount /dev/rz29a /mnt
>     dump 0f - / | ( cd /mnt ; restore rf - )
>     dump 0f - /usr | ( cd /mnt/usr ; restore rf - )
>     cp -p /mnt/genvmunix /mnt/vmunix
>     edit /mnt/etc/fstab: change definition of root and swap device
>     rm /mnt/sbin/swapdefault
>     shutdown -h now
>     boot dkd500.5.0.1010.0
>     Rebuild kernel (just to be clean)
If you carefully manage your single partition, and never manage to do any
really nasty things to it, you'll probably be fine.  After all, we ship the
installation media in very much this format.
Traditionally, disks were much slower and less reliable than they are today.
Since the only way to recover the system, traditionally, was to be able to
boot at least to single user mode, and since traditionally you didn't have a
CDROM drive, and your tape drives weren't really fast or reliable, you did
not want to have lots of read-write activity on your root file system, since
writes are more likely to corrupt things than reads.  So you had a small and
not actively updated root partition, and a possibly large and maybe actively
updated /usr and/or /var in separate partitions (or separate disks).  Lots
of this has been done this way since UNIX was originally developed, and lots
more of it has been done this way since Berkeley UNIX on VAX hardware.
We don't claim to support a running system in the configuration you've made.
That doesn't mean it won't work -- it probably will work fine.  But if you
get into trouble, and you're hosed, and your system won't boot, you'll need
to go back to the installation media and reinstall from scratch, and it's
not going to put things back the way you've got them.
I know of no plans in the foreseeable future to change how things work; in
fact, there are good reasons for doing it the way it is by default, not all
of which apply to workstation configurations with one active user.
Tom
 
 Dr. Thomas P. Blinn, UNIX Software Group, Digital Equipment Corporation
  110 Spit Brook Road, MS ZKO3-2/U20   Nashua, New Hampshire 03062-2698
   Technology Partnership Engineering           Phone:  (603) 884-0646
    Internet: tpb_at_zk3.dec.com           Digital's Easynet: alpha::tpb
     ACM Member: tpblinn_at_acm.org         PC_at_Home: tom_at_felines.mv.net
  Worry kills more people than work because more people worry than work.
      Keep your stick on the ice.        -- Steve Smith ("Red Green")
     My favorite palindrome is: Satan, oscillate my metallic sonatas.
                                         -- Phil Agre, pagre_at_ucsd.edu
     Yesterday it worked / Today it is not working / UNIX is like that 
			-- apologies to Margaret Segall 
  Opinions expressed herein are my own, and do not necessarily represent
  those of my employer or anyone else, living or dead, real or imagined.
 
>From sam_at_cs.stir.ac.uk Wed Feb 25 14:17:52 EST 1998
Subject: Re: Unified / and /usr: good or bad idea?
To: rwgk_at_laplace.csb.yale.edu
Date: Wed, 25 Feb 1998 19:17:07 +0000 (GMT)
From: "Sam Nelson" <Sam.Nelson_at_cs.stir.ac.uk>
You wrote:
| So far everything seems to work just fine. Therefore I was wondering
| why the default is to have the disk artificially fragmented.
| Is there a bad catch when running DEC Unix with unified / and /usr
| partitions?
| 
The original (and I'm going back to Sixth Edition here, late 70s) was that
the 'root' contained just enough to get your system going, from where you
could get it into the state you wanted it for normal operation.  The
'minimal root' idea wins you a lot:
  - given that one error on a disk could screw the whole disk, making the
    root disk as small as possible makes it as resilient as possible;
  - if you do lose your root disk, the 'fundamental recovery medium' root
    disk dump is as small as possible and hence quick to reload
  - almost nothing on the root disk is written to by anyone but superuser,
    so valid recovery dumps can be done relatively rarely
  - if you want to keep a copy of your entire root disk somewhere else on
    another disk, you don't need much disk to do it.
  - etc.
Point is that if you merge root and /usr, it means your root volume is
relatively volatile and needs looked after far more.  That said, it's how
HP-UX (for example) has always been organised, so it's obviously not a
complete disaster.
If I _can_, I like to run with as small a root as I can get away with.  I
don't lose sleep over 2Gb root disks, though...
Sam.
--
Sam Nelson, Comp Sci, Stirling U, FK9 4LA, Scotland         ,->0->M
Email: sam_at_cs.stir.ac.uk    Phone: +44 7050 165499     I->3-+->2->R=->-+->4->O
Office: +44 1786 467443      Fax: +44 1786 464551           `->1->S=->-'
URL:   http://www.cs.stir.ac.uk/~sam                 R$+_at_$+  $_at_smtp$#$2$:$1_at_$2
Received on Thu Feb 26 1998 - 20:59:13 NZDT

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:37 NZDT