SUMMARY: Migrating (slowly) from SunOS 4.1.3 to Solaris 2.3

From: Henrik Schmiediche (henrik@PICARD.TAMU.EDU)
Date: Thu Aug 18 1994 - 00:10:58 CDT

Thanks to all who responded! Not evryone answers are posted below
since some of the answer duplicated themselves.

The following is a summary responde to some of the problems I have
encountered installing my new Solaris 2.3 system. Since I asked a log
of questions I interspersed the answers with my original post ("*"):

*I am running a local network of about 20 or SparcStations, 10 NCD
*terminals and a few other machines. The Sparcs run SunOS 4.1.3 and we
*do use NIS and manually maintained NFS tables. My newest machine is a
*Sparc 20/502MP with a SX graphics card. To my understanding this
*machine requires the use of Solaris 2.3 and is not compatible with
*SunOS 4.1.3. Thus the time has finally come for me to slowly start the
*switch from SunOS 4.1.3 to Solaris 2.3 (or 2.4 by the time it is done).

It was pointed out that a Sun 20/502MP will run fine on a SunOS 4.1.3
machine for users of the TGX graphics card. I knew that, but as far as
I know, the SX graphics card is simply not supported in 4.1.3 - am I

It was also pointed out that you will/might be required to install a
lot of patches and this requires a large /var partition. Question:
Is this also true with Solaris 2.3 Edition II? Has Sun already
included some of the patches in Edition II?

*I would like to ask the people on this newsgroup/mailing list obout
*some of the problmes they faced when in similar circumstances and how
*they proceeded to integrate SunOS 4.1.3 and Solaris 2.x on a single
*network. I have this hunch that this has got to be question that has
*already been asked and I remember something about a SUMMARY archive
*being available.... could somebody point me to it? Thanks.

The summaries are then archived on "" as a gopher, wias
and WWW site. Check out then entry: "SUMMARY: Transition to Solaris 2.3"
which contains a lot of helpful stuff.

*What I am especially interested in is how the file system should be
*organized when you have essentially a SunOS 4.1.3 /usr/local file
*system and a Solaris /usr/local file system...

No clear cut answer. Make sure you install the SunOS 4.1.3
compatibility library and much of the old stuff wil run. THis is what
I did. I started a new "/usr/local" called "/opt/local" which contains
Solaris specific stuff. The "/usr/local" stuff is mounted from the
SunOS 4.1.3 machine and the search paths are set up to first try the
Solaris "/opt/local" directory and then then "/usr/local"
directory. Some stuff in user local will not work, but as you discover
it you can port it to "/opt/local". (

Other suggestions:
As far as making applications available on Solaris and SunOS
platforms, I use the (SUN-developed) appmap program. This is a shell
script with accompanying config file which enables the correct app
to be selected depending on the OS (and other factors, too).
grabbed "appmap" from the SUNSITE archive at unc. (

...with the automounter. The /usr/local bit is the easy part, since
you can just automount it from one machine or another, depending on
which version of binaries you want to run (remember that you'll still
be able to run some SunOS 4.1.3 binaries on your Solaris machines
too). The fun bit comes when you try to consider /opt, since Sun
haven't really considered how you're supposed to get everything
working. I've got /opt under control of the automounter too, and
automount the various additional packages from different servers as
appropriate. It works well in practise, but it's a pain installing
software, because the installation scripts barf when they discover
they can't write to /opt. Little do they know that even though
/opt/<pkg> doesn't exist, it will magically appear when they try to
access it... Make /var *big*. You've a lot of patch installing to do,
and that's where they go. (

*The other issue is NIS. Is it difficult to make the Solaris machine a
*client of the SunOS 4.1.3 NIS master (I had trouble finding a relevant
*section in the answer book that talks about this - it tells me how to
*set up NIS+ masters & clients and talks about NIS compatibility from
*the Server point of view, but perhaps I missed something) or
*do I make the Solaris machine the NIS master and the SunOS 4.1.3
*machine the clients? Either one is ok with me. I ultimately (1yr
*or so) would like all our machines to be Solaris 2.x

It was easy to make the Solaris machine a NIS client. There are two
ways of doing it:

   1) run /usr/sbin/sys-unconfig and reboot & reconfigure

   2) the manual method:

       The manual method is also relatively easy (also have a look in
       /etc/init.d/rpc where this normally starts):

    a. create /etc/defaultdomain with the domainname in in (same as 4.x)
    b. run domainname `cat /etc/defaultdomain`
    c. cp /etc/nsswitch.nis /etc/nsswitch.conf
    d. /etc/init.d/rpc stop
    e. /etc/init.d/rpc start

      and you should nowe be using NIS as a client. (

I did #1, can't vouch for #2, but it is good to know. My machine is
now a NIS client.

More stuff:

  First off, you don't get NIS software bundled with Solaris 2.3 - if
  you want it, you'll have to get it from Sun (NISKit, or something;
  haven't tried it). This means your default options are: NIS+ in
  YP-compatible mode, or full NIS+. Welcome to hell. You *can* make a
  4.1.3 machine the NIS+ master, since Sun do supply that, but I haven't
  tried it. I *strongly* recommend running NIS+ by itself on one Solaris
  machine and getting the hang of it before attempting to spread it
  out. Don't go for multilevel domains, because there are still bugs in
  this. NIS+ is ok when it works, and a real pig when it doesn't.

*I did set up the DNS name service almost immediately (since I was not
*famialar with NIS+) and I noticed that Solaris takes about 2 seconds
*to located a machine which is not local. Under SunOS 4.1.3 it usually
*did not take more then half a second. Anyone know why DNS is so slow?

This problem dissapeared when I made my machine a NIS client. Here is
one response for a possible cause:

  Check /etc/nsswitch.conf. This file sets the order that the resolver
  checks /etc/hosts, DNS and NIS. Perhaps your machine is waiting for
  NIS to time out. Also check /etc/resolv.conf to make sure that a
  local nameserver is first. ( )

*The printer setup in Solaris sure is nice, but I can't get
*my Solaris machine to communicate with the remote printer.
*I went through the trouble shooting section and eveything seems
*to be fine. "lpstat -p" says:
*printer hp1 faulted. enabled since Thu Aug 11 08:59:18 CDT 1994. available.
* system not responding
*printer sp1 faulted. enabled since Thu Aug 11 09:04:50 CDT 1994. available.
* system not responding
*The log entries say:
*08/11/94 12:33 p 169 <none> Starting.
*08/11/94 12:33 p 169 <none> Starting lpNetParent.
*08/11/94 12:33 p 169 <none> Initialized & Polling.
*08/11/94 12:33 p 169 <none> Started child for lore, pid = 178
*08/11/94 12:33 p 169 <none> Started child for picard, pid = 180
*08/11/94 12:33 c 178 lore NAKed by remote lpd (108)
*08/11/94 12:33 c 180 picard NAKed by remote lpd (112)
*08/11/94 12:43 c 178 lore waiting for remote queue to be enabled
*08/11/94 12:43 c 180 picard waiting for remote queue to be enabled
*The remote queues are enabled (as far as I know). The remote printer
*to say "no deamon present" and "lpc restart sp1" don't fix that (as
*the answerbook trouble shooting section implies), but I never had any
*problems printing. Anyone know what in the world is a "NAKed" and how
*does one fix this problem?

NAKed is an abbreviation for "not acknowledged". I have not fixed this
problem but I have not yet tried eveything suggested below., says:

  I think this problem is caused by not having the host's IP numbers in the
  /etc/hosts file. You must have the print client host IP in the print
  servers /etc/hosts file, and also the print server's IP number in the
  print clients /etc/hosts file. It is not sufficient to have them in the
  NIS(+). says:

  As far as your printers, I will assume from your little lpstat -p output
  that these printers are remote, ip type printers. The enclosed script
  needs to be edited so that PRINTER is set to some real device on your net,
  and it works here like a champ.

        lpsystem -t bsd $PRINTER
        lpadmin -p $PRINTER -s $PRINTER!$PRINTER -I simple,postscript
        enable $PRINTER
        accept $PRINTER
        lpadmin -d $PRINTER says:
  Have you created /etc/hosts.lpd? If not see the manpage for lpd(8):
  Access control is provided by two means. First, all
  requests must come from one of the machines listed in either
  the file /etc/hosts.equiv or /etc/hosts.lpd. says:
  For the printer, try deleting it, and using admintool to add it.
  (Repeat after me: "Admintool is my friend. Admintool is always right.
  If I ever get the time I'm going to learn enough about the system that
  I'll never have to use admintool again.") Eventually you'll want to
  use the script interfaces, which will make life much easier as you
  upgrade more machines. To add a new remote printer, you just say
  something like
    # lpsystem -t bsd host
    # lpadmin -I 'any' -p lw1 -D 'Room 165 Apple Laserwriter' -s sag4
    # enable lw1 ; accept lw1

Note: I did use admin tool. I have not fixed this problem, but I have
not tried adding the printer manually.
*Finally, I am currently running straight X11R5(pl26) on the network
*with the intention of moving to X11R6 soon. I already have X11R6 up
*and running on the Solaris machine, but I am currently using
*openwindows 3.3 since the 'answerbook" does not seem to like X11R6
*(don't work). Even if I run openwin with the X11R6 server I cannot get
*the answerbook to work. Anyone know how to get the answerbook to work
*with X11R5/6? What do people prefer, openwin or X11R5/6? I hear
*openwin is a performance hog - is that true? Certainly seems sluggish
*to me at times.. but perhaps I've been biased.

Answerbook uses display PostScript which is not available in X11R6.
You can write a wrapper so that answerbook calls GhostView to display
the answerbook files (anyone done this?).

You can speed up OpenWin by using Sun's "dxlib stuff" which is an xlib
replacement that write directly to the frame buffer. I mam not sure
this helps if you want to use X11R6. Anyone know when Sun's
Openwindows X11R6 server will come out?


Thanks again to all who answered. Appreciate it.

   - Henrik Schmiediche

Henrik Schmiediche, Dept. of Statistics, Texas A&M, College Station, TX 77843
E-mail:  |  Tel: (409) 862-1764   |  Fax: (409) 845-3144
Finger for pgp 2.6 key, fingerprint: E867 D9DB 9616 5DAC  0F67 FE98 77FE 8583

This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:09:08 CDT