Subject: named eating 100% cpu
To: None <current-users@netbsd.org>
From: Anthony Mallet <anthony.mallet@useless-ficus.net>
List: current-users
Date: 08/25/2003 22:01:19
Hi,

Using 1.6W dated Fri Aug 15 22:53:08 CEST 2003, I noticed that named
spirals down into and endless loop, eating 100% cpu when one interface on
which it is listening goes down and then up again. I wondered if it was
an already known problem or if I should dig further to find some more
precise information...

Another problem which appeared at the same time concerns ssh tunnels or 
ssh shells which I create from outside my private network to my
gateway. The connection is a DSL link which is periodically reset by the
provider. When the link comes up again (after 10 seconds or something),
all the connections that I established from the outside are closed,
whereas the connections that I had established from inside my network to
the internet are kept alive.

I have no precise idea regarding where to start to debug those two
problems -- if these are problems and not 'features' :). So if you
do have some hints... I'd be glad :)

If it has some importance, the interface which causes problems is a ppp
link, driven by pppd and rp-pppoe.