Hi Rick,
Reference your note (copied below).
Here, at the IBM UK AIX Systems Support Centre, we use a single
RS-232 terminal connected to a terminal server which is in turn
connected to several RISC System/6000 production host console ports.
We use a simple script ("cycle-console" see below) and cron to switch console
to a new console log file (in /var/log) each day (this script is also run at
boot time by our rc.local). Our "skulker" deletes console log files that
are older than 7 days.
This is a local tool and I can offer no "warranty" on it
but I hope this helps you.
--
cheers
paul finger [EMAIL PROTECTED]
-- cycle-console follows (in shar format) --
#!/bin/sh
# shar: Shell Archiver (v1.22)
#
# Run the following text with /bin/sh to create:
# cycle-console
#
sed 's/^X//' << 'SHAR_EOF' > cycle-console &&
X#! /bin/sh -
X#
X# @(#)cycle-console 1.22 (ASSC) 6/13/94
X# /vol/src/generic/usr/local/sbin/cycle-console/SCCS/s.cycle-console
X#
X# NAME cycle-console
X# AUTHOR Paul Blackburn <[EMAIL PROTECTED]>
X# PURPOSE switch console to a different log file
X# USAGE Typically invoked via a daily crontab job or /etc/rc.local
X# sample cron entry:
X# 1 0 * * * /usr/local/sbin/cycle-console
X#
X# DISCLAIMER
X#
X# The information in this document has not been submitted to any formal
X# IBM testing and is distributed on an AS IS basis without any warranty
X# either expressed or implied. The use of this information or the
X# implementation of these techniques is a customer responsibility and
X# depends on the customer's operational environment. While each item
X# may have been reviewed by IBM for accuracy in a specific situation,
X# there is no guarantee that the same or similar results will be
X# obtained elsewhere. Customers attempting to adapt these techniques to
X# their own environments do so at their own risk.
X
X
XCMD=`basename ${0}`
XOLDCONSOLE=`lscons`
X
Xif [ ! -d /var/log ]; then
X mkdir /var/log
Xfi
X
X# We cycle between /dev/tty0 and /var/log/console* to enable the system
X# to recover a console on /dev/tty0 if there is a problem with
X# /var/log/console* (see swcons man page for detail).
X
X/etc/swcons /dev/tty0 >/dev/null 2>&1
X
X# We append hour minute second to enable cycle-console to be run anytime
X
XX=/var/log/console.`date '+''%h''%dat''%H''%M:''%S'`
X/etc/swcons ${X} >/dev/null 2>&1
Xchmod 644 ${X}
X
Xcase ${OLDCONSOLE} in
X /var/log/console* )
X <${OLDCONSOLE} compress >/tmp/${CMD}.$$
X rm ${OLDCONSOLE}
X mv /tmp/${CMD}.$$ ${OLDCONSOLE}.Z
X ;;
Xesac
Xrm /var/log/console 2>/dev/null
Xln ${X} /var/log/console
X
XSTAMP=`date '+''%H''%M'':%S on %a %d %h %y'`
Xecho "# console output for `hostname`" >/dev/console
Xecho "# commencing at ${STAMP}" >/dev/console
Xecho "# achieved using /usr/local/sbin/${CMD}" >/dev/console
Xecho >/dev/console
SHAR_EOF
chmod 0555 cycle-console || echo "restore of cycle-console fails"
exit 0
-- end of cycle-console --
-- copied note follows --
On Mon, 13 Jun 1994 10:09:17 -0400 (EDT)
[EMAIL PROTECTED] wrote:
> We just wasted a bunch of our time and Transarc's with a fiendishly
> simple problem.
>
> We have removed the fancy graphics adapters from our RS6000 AFS
> servers and are using RS-232 terminals as consoles. On one of our
> servers, the RS-232 connector came halfway unplugged - "receive" and
> "transmit" were still connected, but "DTR" was not.
>
> AIX apparently ignores DTR until it goes multiuser, so it looked as
> though the console was working. However, every process which tried to
> write to the console (including the AFS "fileserver" process) blocked
> quite thoroughly when it couldn't write to the console!
>
> This took embarrassingly long to find. As a matter of fact, I'm amazed
> that I'm willing to admit to this nonsense in public!
>
> A hardware solution to this is to securely fasten to the console
> RS-232 cable an adapter which passes "receive" and "transmit", but
> jumpers the modem signals so that the RS6000 serial port always thinks
> it has a working terminal attached. Simply fastening the cable to the
> terminal is not sufficient since taking the terminal offline, turning
> its power off, or unplugging its power would still cause the server to
> fail. Unfortunately, so would unplugging the RS6000 end of the cable
> even with the adapter.
>
> Perhaps there is a software solution. If so, I would use both.
>
> Caveat system administrator.
>
> -Rick
>
> --
> |Rick Cochran 607-255-7223|
> |Cornell Materials Science Center [EMAIL PROTECTED]|
> |E20 Clark Hall, Ithaca, N.Y. 14853 cornell!msc.cornell.edu!rick|
