[cobalt-users] Re: Help! The case of the disapearing remote Raq3...

["David Thacker" <Cobalt@xxxxxxxxxxxxxx>]

> Recently, our Raq3 has developed a habit of "disappearing" from the net
(no
> services, no pinging, no nuttin'!) for about 45 minutes at a time (It's
> hundreds of miles away in a colo place, so for all I know it's out getting
> coffee).
>
> Regardless, the logs show that swatch and other cron programs are running
> faithfully every 15 minutes during the outage, but absolutely NO network
> traffic is measured. There is no record of the services shutting down or
> restarting either -- the machine just stops answering. Dead silence for
> almost an hour, and then suddenly, it shows up again, happy as a a clam
and
> answering all its ports as normal.
>
> ANY clues as to what might cause this behavior?? It's driving me -- and my
> frustrated clients -- up the wall!


Greetings,

The dirty little secret is that this behaviour just sometimes happens on the
RaQ3s.  If you scour the archives back a couple years, you'll find similar
posts about this problem, mine included.

Trying to resolve it with Cobalt support was unsuccessful.  I finally did
find a cure that fixed my RaQ3 though: simply making it ping out to the
gateway on a regular basis was enough to keep the machine visible on the
net.  Here's what I use:

nohup ping -i 65 -n gate.way.ip.address &

Su to root, then enter this command, which tells your RaQ to ping the
gateway once every 65 seconds running as a background process.

Actually powering down the server (not just rebooting) does seem to make it
happier too.

My colo guy came to the conclusion that it was some weird antisocial
behaviour between the onboard Intel ethernet port and certain Cisco 27xx
switches.  He figured it might be possible to solve the problem by rolling a
newer Intel ethernet driver into the RaQ3 kernel, but we never went that far
to see if that would do the trick.

Good luck,

dAvid tHacker
Thacker Network Technologies Inc.
Cobalt@xxxxxxxxxxxxxx