AIX

Hi All,

I am working on a server which has an /etc/syslog.conf file with the following entries (example):

local6.debug /dplogs/uss1udp001.log rotate size 1500m time 1d files 14
local7.debug /dplogs/uss1udp002.log rotate size 1500m time 1d files 14

-----------------------------
Basically this config file is taking system messages from Tivoli, which are at the 'debug' level, and sending them to logfiles in the /dplogs directory.
It rotates 14 of them based on 1 day or 1500m size.

This is all well and good. However, what I need to do is execute a UNIX script every time an entry gets dropped into one of these files.

Does anyone know a way that you can use /etc/syslog.conf to, instead of dropping the message into a directory, execute a file instead?

The entry would be something like (in this example)

local6.debug /dplogs/runit # every time a debug entry, kick off script, no matter when


Possible? Anyone done something like this before. Thanks much for any advice.

jeffpas

The only thing I can think of to do is run the script via crontab, every minute of every day, such as this:

* * * * * /usr/bin/myscript

But this project manager wants the script to run 'realtime', meaning every time a log message writes from syslogd.

There does not seem to be any way to trigger a script to run through syslogd..........

I have not tried this, it is not tested, but an idea you could pursue:

syslogd can send messages either to a file (this is the common case), to another host or to a users terminal. For instance: to send all mail and daemon messages of priority crit and above to user xxx enter the following line into /etc/syslog.conf:

Code:

mail,daemon.crit xxx

You could probably create an user account, have it run some script which waits for terminal input and does something according to the sort of input it receives. The create an entry in /etc/syslog.conf to have this user receive all the messages you want to track.

i don't know if you could specify several destinations so maybe your script will have to do the log writing syslog normally does by itself.

In any case i would be intereseted to hear your solution.

bakunin

bakunin,

This appears to be a question that will never die.

I have since found a way (by creating a daemon that checks the logfile for any changes, every 20 seconds or so) to rig a solution, but everyone involved in the project keeps bringing up this trigger mechanism again and again.
It seems they are determined that I do it that way.

All I can say is one of the techs played around with this briefly and suggested creating a fifo pipe file. This is something I have never done, or been told to do, in about a decade and a half of UNIXing. Sorry, it just has never come up.

Apparently it works something like this:

#!/usr/bin/sh
exec 0 < log_pipe #log_pipe is a pipe made using the mkfifo cmd
while :
do
while read LINE
sleep 1
done


The only thing I can possibly understand about a fifo file is that it simply sits with the contents of one action and waits for another, such as:

ls | pg

you can mkfifo pipe_listing then:

ls > pipe_listing

then pg < pipe_listing sometime later.

Any of this make any sense?

Everyone says that they can't understand why I am not going htis route, but no one can say how................can't figure it out for anything................