Ethernet interface status detection

Hi all,

Thoughts on a clean and efficient way to do this would be appreciated
thanks.

We have an issue where our linux gateway resets each night. However,
last night the ISP did not hand out a IP address or some other crucial
part of information. As such during the reboot the eth<X> did not come
back up. We had to wait until a good point to reset it, as it also
provides NAT access between arious networks. What would you all
recommend running on say a cron job (open to suggestions), to detect
if an interface is not correctly connected to the internet, so that if
could be brought down and back up. As i use this for remote access all
the time, up and down no matter what, is as a preference not a nice
option.

Please advise.

David

P.S The Linux version is Gentoo, with a full custom kernel build for
this purpose. However, i guess whatever variant of Linux if its
availale then i should be able to setup.

Thanks again in advance for any ideas.

David <(E-Mail Removed)> wrote:

> We have an issue where our linux gateway resets each night. However,
> last night the ISP did not hand out a IP address or some other crucial
> part of information. As such during the reboot the eth<X> did not come
> back up. We had to wait until a good point to reset it, as it also
> provides NAT access between arious networks. What would you all
> recommend running on say a cron job (open to suggestions), to detect
> if an interface is not correctly connected to the internet, so that if
> could be brought down and back up. As i use this for remote access all
> the time, up and down no matter what, is as a preference not a nice
> option.
>

I used to do various checks in a script and I replaced this with a much
simpler ping to google.com - if this succeeds then my DNS is OK, my route
is setup, and I can talk to the net. I realise that this could fail if
google stops responding to ping echo-requests (unlikely) -- more than often
there is as you say a slight misconfig and all that is needed is a quick
ifup/ifdown. So I have the following in which I'm using ppp0 and I've
expanded it for clarity. I'm assuming that your firewall allows outbound
pings and incoming replies..

/usr/local/bin/testremotenet:

#!/bin/sh
semfile=/var/run/remote-net-up
rm -f $semfile
ping -c4 www.google.com &>/dev/null
if [ $? -ne 0 ]
then
/sbin/ifdown eth0
/bin/sleep 15
/sbin/ifup eth0
if [ $? -ne 0 ]
then
mail -s "eth0 dead at $(date -R)" \ root@localhost </dev/null
exit 1
fi
fi
touch $semfile
exit 0

and in root's crontab:

0,15,30,45 * * * * /usr/local/bin/testremotenet &>/dev/null

will run this every 15 mins. You can drop in extra bits between ifup and
ifdown if you have a trick to get the interface back up.. or maybe loop
this bit to have a few more goes at getting it going before dying. Good
luck.

--
Jamin @ Home: Chester UK -<(E-Mail Removed)>