On Sat, Feb 25, 2006 at 03:33:28PM -0800, Peter Bako wrote:
> I have a weird problem I cannot find a solution to. I've written a small
> script (attached below) that I put on the dozen or so systems that I
> maintain for friends and clients, that daily sends some basic information to
> my web server. This data is then stored in a MySQL database and viewed via
> another script. All the systems are running OpenBSD version 3.5 to 3.8, and
> the one in question here is 3.8.
>
> The problem is this. On one remote system (identical in every respect to
> about 8 others out there), the script when executed manually (either as root
> or as a non-privileged user) runs normally and uploads its data as it
> should. However when the cron job hits at midnight the script always fails
> and without any error message that I can get. As you can see the script is
> quite simple, the only active component is a call to CURL which hits a
> specific address. The local log entry lists my error message but $result is
> always empty so I have no specific error to go by. By looking through the
> logs of my own web server at the same time that the local log entry is made,
> I know that the connection to my system is never established.
>
> Here is the script:
> ----------------------
> #!/bin/sh
> name=`uname -n`
> ip=`ifconfig sis0 | grep 'inet ' | awk '{ print $2 }'`
> space=`df | tail -1 | awk '{ print $4 }'`
> ver=`uname -r`
>
> data="http://xxx.yyy.com/fw/fwin.php?NAME=$name&IP=$ip&FREE=$space&VER=$ver";
>
> result=`/usr/local/bin/curl -s $data`
> case $result in
> good)
> `logger Info sucessfully logged!`
> exit 0
> ;;
>
> *)
> `logger Unable to log system info! Error: $result`
> exit 1
> ;;
> esac
> ---------------------
> The cron job that launches it is added to root's crontab (crontab -u root
> -e) and looks like this:
> ---------------------
> @daily /usr/local/fwreport
> ---------------------
>
> I've tried leaving the -s flag off of the CURL call to get some kind of an
> error out, but whatever might come back does not make it out to the $result
> variable. Again this identical script works on over a dozen other systems,
> most totally identical to this unit down to the hardware and OS version, so
> it has to be more or less correct.
>
> Any suggestion, ideas, etc. are appreciated.
Aside from environment variables, as Reid suggested (then again, the
most likely problem is curl not being in PATH, and you evaded that
problem by using a full pathname), I'd not immediately come up with
something. However, a little something that may help with debugging:
result=`/usr/local/bin/curl -s $data 2>&1 || \
echo "curl failed: exit status $?"`
Which is likely to yield you some more debugging information.
Also, I'd suggest using https for this sort of thing, if you feel the
need to use http at all, but that's a different point.
Finally, why do you ``-quote the logger command?
Joachim
