[rescue] Questions about SunOS 4.1.4

Sandwich Maker adh at an.bradford.ma.us
Sun Aug 10 00:50:51 CDT 2003


"From: "Curtis H. Wilbar Jr." <rescue at hawkmountain.net>
"
">From: William Barnett-Lewis <wlewis at mailbag.com>
">
">> >From: William Barnett-Lewis <wlewis at mailbag.com>
">> >
">> >I'm getting my hands on an IPX and was thinking silly thoughts of using
">> >this copy of 4.1.4 I have sitting here. This is just for the fun of it
">> >and to perhaps play with sunview.
">> >
">> >Did this version still include a CC? And which version of gcc will
">> >compile for this old a beast? I figure once I get gcc up, then I'm ok
">> >for the rest of the necessities (emacs, perl, etc).
">> 
">> It includes cc.
">
">Helpful.

two flavors, /usr/bin/cc and /usr/5bin/cc.

">> The newest version of gcc should work to the best of my knowledge
">> (I don't believe they have removed support for SunOS 4.X).
">
">Well, that's what I thought but I wasn't sure if I would need to start
">with something like CC to GCC 1.4 and then up the food chain to 2.95 or
">3.x. 
"
"I seem to recall compiling (foggy memory) gcc 2.something using cc.
"(The actualy "procedure" is compile gcc with cc, then recompile
"gcc with gcc, seem to recall a 3rd compile too... but it has been
"awhile).

yup.  the idea was to converge on a consistent gcc-compiled work.

">> Other than patching for security, SunOS 4.X is a very stable and quick
">> OS.
">
">I found an FTP site with all the known patches... That should be
">interesting to install. OTOH, better this than anything from Redmond...

they are; there's no patch-mgmt system for it.

and despite their earlier intransigence, there -are- y2k patches for
it.

i was successful in dividing the patches into three groups: kernel,
openwin, and everything else.  i made empty directory trees on my
patch server, placed the patch files into their proper places, and
used rdist.  i could do the kernel files any time, openwin any time ow
wasn't running, but the others...  iirc, i discovered that rdist
worked by actually copying the file to a temp name, then renaming it
when the copy was complete.  if the binary was running, the disk file
was locked and the mv would fail on the unlink.  if i took note of the
failures, i could do something like rename the running bins [that was
allowed], then rdist, then reboot, then rm renamed old bins.

iirc no systems crashed when i did this, but i didn't hit the problem
often.  otoh i hit it on ~60 systems every time i did...

"I really like SunOS 4.X.  I like BSD, although I'm quite used to Solaris
"now.

i got quite fond of sunos4 too.

"4.1.3 (Solaris 1.1) was always my personal favorite.  I never really
"saw much of a reason for 4.1.3_U1 (Solaris 1.1.1), 4.1.3_U1 Rev B
"(Solaris 1.1.1 Version B), or 4.1.4 (Solaris 1.1.2).  There was
"integration of fixes, as well as support for new architectures 
"(Classic, LX, Sparcstation 5).  

4.1.4 also included support for ross modules and stabilized multiproc
operation on 10s and 20s even with ti modules.  [i worked at a shop
that did heavy chip design and sim, around '95-6.  we had a 10 and a
20 up for over three months with dual supersparcs when i left.]

"When I was the CTO of a regional ISP, we ran Solaris 1.1 for a long
"time (When we migrated, we migrated to Solaris 2.6, and I think 7
"had just come out).  We also were running lots of SS2, Sparcstation 10,
"and a few Sparcstation 20.  We had over 10 SS2 boxes running loads of
"web domains... we migrated them all to an E450 running Solaris 2.6.

my most satisfying contract to date involved taking a group running
mostly sunos4 [10s and 20s] and disks humg off the systems and
cross-mounted willy-nilly, but also three ultras running s2.6, and
updating everyone to s2.6.  during the job my boss scared up another
20 with a ssa100, so i also consolidated the nfs spiderweb...  that
took work.  they had multiple copies, fragmented everywhere.

i moved -everything- special to the server, then created a custom
jumpstart that reinstalled all the systems with a fresh s2.6 over lunch
on a friday.  in the process, i swapped the server ss20 for the ultra1
on one of the desks, and straightened the mess of tool installations
[multiple copies again], and fixed a shopping list of user issues that
had been impossible in the previous tangle...

the group walked in, logged in, and went back to work.  that was cool.

and when one of the systems crashed messily, little was lost - the
user could even move to another system temporarily - and when the
cause didn't yield quickly to inspection, we just re-jumpstarted it
and it was back to operation in 15m.

"Even created a custom patching tool for SunOS 4.1.3 that applied all
"the patches including modifying the rc files to install the components
"that needed to be installed in single user... which then restored the
"rc files after patching and rebooted again....  made patching 4.1.3
"a breeze.  Drastically improved setup time for a new system.

many patches don't need to be installed in single user, but the ones
that should, -really- should.
--
you might be interested in three little programs i still have lying
around since then: a warm-boot accelerator for ss2 and ipx, a kernel
tuner for sunos4 [using early cockroft rules], and a disk
partition-map calculator.  should i post them?
________________________________________________________________________
Andrew Hay                                  the genius nature
internet rambler                            is to see what all have seen
adh at an.bradford.ma.us                       and think what none thought



More information about the rescue mailing list