My apologies for the delay of this belated second SUMMARY.
The problem:
Our SPARC Server 1000, Solaris2.3, with the recommended patches, and
supporting many users, dumped core in /.
We installed the Solaris2.4 11/94 upgrade, including recommended
patches. All new users were denied access to the machine. Solaris
2.4 did not dump core, but the machine was virtually inaccessible.
The solution:
The useradd program turned out to be the cause of the problem. (We
have only a Wyse50 terminal attached to the SPARC 1000, and accounts
were being added by an "accounts administrator" using sudo useradd.)
As soon as I edited /etc/passwd and created the user directories by
hand, and ran pwconv, the problem went away.
Digest follows, with many thanks to Sun Managers.
----------
From: Gavin Maltby <gavin@DURBAN.VECTOR.CO.ZA>
Date: Sat, 14 Jan 1995 14:45:36 +0200
Hi,
Do your users have capitals in their login names? If so, especially
if the initial letter is a capital, then the login process sometimes
(I'm not sure of the rules) assumes you have a brain-dead terminal with
no lowercase and converts to lower case for you. This bit me recently
in setting up PPP stuff.
Otherwise, one can figure what is happening by trussing various
processes. e.g, wait till login: is displayed, find the PID
and truss -f -o /tmp/trace -p <PID> and then analyse the trace
file to see what was read when the user typed what. I solved
my problem by using a temporary login "wrapper":
mv /bin/login /bin/login.real
cat <<EOF
#!/bin/sh
truss -o /tmp/trace -f /bin/login.real $*
EOF > /bin/login
chmod 4755 /bin/login
This won't actually allow login (utmp entries not around) but writes
adequate trace files for debugging. Be sure to keep a root shell open
on the console!
In the trace file you'll see login write "login:" to STDOUT
and then read from STDIN (ie, write(1,...) and read (0,...) )
and you can check if that lines up with what was really typed.
If things don't line up, it may not be the fault of Solaris.
I had a customer with a weirdly set up Kermit/telnet that
caused related problems.
Cheers
Gavin
--------
From: Santithorn Bunchua <keh@au.ac.th>
Date: Sun, 15 Jan 1995 00:02:19 +0700 (GMT)
Please take a look at the patch 101855-01 .. I installed this
on my 2.3 and the core disappears.
---------
From: Geert Jan de Groot <GeertJan.deGroot@ripe.net>
Date: Mon, 16 Jan 1995 11:49:54 +0100
How about analyzing the core file, eventually helped by a symbol table
to be supplied by SUN? At least SUN must to be able to analyze these.
This is what you get with binary-only systems - it's why I stopped
using SUN when they switched to solaris (with sunos, you could usually
take the 4.3BSD source & recompile)
Swapping hardware will likely not fix your problem, though playing
around with adb might.
Geert Jan
-----------
From: mark@ccifl.com (Mark Sigler)
Date: Mon, 16 Jan 1995 09:54:20 -0500
George,
I experienced some problems after installing
Solaris2.4 that were related to the serial driver.
The symptom was difficulty in login/password
due to faulty keyboard input (unrelated to your problem).
I reinstalled, using the Entire Distribution option
and the problem was gone (???... go figure).
There is a patch that is on the distribution media
that is a in-kernel rlogind/telnetd for 1000/2000
servers to support large mulit-user access.
I can't recall the patch number, but check the README's.
I can only hope that you have a support contract
(I assume you do, since they're replacing hardware)
or can persuade your vendor to harass Sun Tech Support
on a daily basis. You need to have your call escalated to
SMCC Engineering.
I'll put out my feelers...
Please keep me posted, as I anticipate supporting several
of my customers through this upgrade process.
Thanks,
Mark Sigler
Systems Architect
Open Systems Technologies
(305) 968-8591 xt. 113 (voice-mail)
(305) 896-4386 (pager)
Mark.Sigler@cciFL.Com (e-mail)
----------
From: Rik Harris <Rik.Harris@fulcrum.com.au>
Date: Mon, 16 Jan 1995 07:55:34 +1100
It's a bit of a long shot, but we had problems with host aliases for
"loghost" under Solaris 2.3. Apparently they are no longer legal. I
had a quick look for the bugid/srdb numbers, but I couldn't find them.
If you search sunsolve for loghost you might be able to get them.
good luck,
rik.
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:15 CDT