Help! Unix greenie with scripting question ( I think) 1

[cschuette]

Hi all,
I'm an automotive designer knowing almost nothing about Unix except that I am now working on a Unix box. We are running a windows-like environment where we can open xterminal windows to run Unix commands. Here's what I need to do...on certain days we have network and software upgrades scheduled at certain times. It is imperative that my workstation is logged off the network by this set time on certain days or I may lose valuable data. Is there a way to write a script that would log my workstation out at a set date and time? I think that this would be a foolproof way to ensure that my workstation is logged out when these updates occur. (Can't always count on my memory unfortunately, especially when I'm really busy). Is there a way to do something such as this or am I barking up the wrong tree? I'm sorry that I can't really provide better information at this point as I am quite new to Unix as previously noted. Any help or opinions would be greatly appreciated. Thanks in advance.

Chad
cschuett@visteon.com

 

[rarnold1]

write a script called killme.sh


#!/usr/bin/ksh
## kills process launching this script
PROCESS_ID=$1 # uses 1st passed parameter # as the value for PROCESS_ID
KILL_HOUR=18 # choose your hour
KILL_MIN=05 # choose minute of hour
END="NO" # set a flag

until [[ ${END} = "YES" ]]
do
if [[ `date +%H` = $KILL_HOUR && `date +%M = ${KILL_MIN} ]] ; then
kill ${PROCESS_ID}
END="YES"
else
sleep 20 # pause 20 sec
# (could be up to 60)
fi
done



then, in the .profile file for the user, end the file with the command
"bg killme.sh $$"
this will put the killme.sh script in the background and pass the process number as the first parameter and kill you when the appropriate time arrives. This will happen
just as you login without having you remember
to do it.

 

[KenCunningham]

Just a word of caution - this script is excellent as long as your system management team is using the system clock (perhaps in a cron job) as their guide to timing the upgrades or whatever. However, if not, you may find that people tend to use their watches to decide when to kick these things off - invariably these are not in sync with the system's time. Therefore, if the sysadmin's watch happens to be 5 minutes behind the system time, fine, but if it's five minutes ahead, oops. I think I'd build in a little leeway in your timing in an attempt to counteract this human/machine interface effect as it's not known in the trade.

OK, call me Cassandra if you will!!

 

[mrregan]

Hi,
I believe you may be approaching this backwards. If users need to be off the system during updates, the administrators should take the system down to single user mode so EVERYONE is warned to get off before their processes are killed, and no one can accidentally log back in before the work is complete. Hope this helps.

 

[toolkit]

If you intend to use kill signals, you must make that your processes and data handle kill signals gracefully. Otherwise you will lose data. With shell scripts, you can use [tt]trap[/tt] to handle signals. With compiled binaries, you should study the documentation/source code to see what signal handlers have been defined.

Also, rather than use killme.sh, a simpler solution may be to assign a value to [tt]TIMEOUT[/tt] (for bourne shell) or [tt]TMOUT[/tt] (for korn shell) in your [tt].login[/tt] script. This will cause your login session to expire if no keyboard input has been recieved after the interval specified in this variable. Try this out using:
[tt]
$ env TMOUT=1 ksh
$
shell will timeout in 60 seconds due to inactivity
ksh: timed out waiting for input
$
[/tt]
[The blue text indicates the second ksh invoked, and immediately timed out]

Cheers, Neil