We have 1 Linux 1.2.8 Box for pri DNS, pri radius and sendmail (P-90 32 megs)
1 Linux 1.2.8 Box for sec DNS, sec radius and news (P-90 128 megs)
1 linux 1.2.8 Box for shell login (NFS Mounted to sendmail
/var/spool/mail on first box) and personal web pages (486-80 32 megs)
3 PMe30's for dialin 28.8/ISDN
1 irx-114 to the net
1 irx-114 to a customer
As IP traffic is VERY high on our network, we use an SMC Tiger Switch
(ether Switch) <GREAT piece of hardware that does RIP>. Each IP Box has
its own rip routed ether port on the switch.
ALL machines are on the same Class "C" (as are all assigned addresses) ALL
PMe30's are set to Broadcast, Listen (on) in both global settings and ether0.
The irx-114 going to the net is set to Broadcast, listen (on) on ether0,
listen under global and RIPsend under the S1 setting (T-1 Link). All linux
boxes do NOT have any routing turned on.
We have had a bunch of problems with our high cap circuits during the
last few weeks. Bell has been working on them, BUT.....
When the T to the net goes down dialin users CAN NOT GET LOGGED IN. Radius
seems very slow and comes back with invalid login. This will continue as
long as the T is down. When we try manual login via the modem/PM we can't
get in ourselves. I don't know for sure if its EVEN getting to the
/etc/raddb/users file (on the first linux box or even trys sec radius on the
second linux box)....I don't think it is. Messages log shows unsucessful
login userid for all logins....Until the T comes back up.
It looks like something is driving the local ether bonkers when the T to the
net goes down (I speculate - I do NOT know this for fact - I just don't know
what else could be happening???).......And radius can't seem to get through
and just gives up. The log will show unsucessful login for each user trying
to get on when the T to the net is down..... I thought sendmail might have a
bunch of mail in the que (with all the .gov and .mil's being down), and was
causing a problem with UDP (DNS lookups off site)) and I tried killing the
sendmail process....Didn't change a thing.
I think (UDP) is working OK as we can telnet (TCP) to names on the local
ethernet network.
When the T works....everything is fine, and has been fine for over a
year (We have had very good luck with our T <grin>)
I have just about pulled all my hair out the past few days trying to figgure
this one out (getting VERY bald <grin>)....as it just compounds our problem
with the link being down to the net. I can't even leave a message to all
explaining that Bell is working on the T....because they can't get logged in
to read the message, our phones will ring off the hook.....
Any help would be greatly appreciated, as this one has driven us crazy
and I don't know when Bell is going to get our T stablized.
It is probably something obvious and stupid that we have missed....but
P L E A S E H E L P ! ! ! ! ! ! !
Frustrated in Virginia....
Steve Haynes steve@exis.net
Presiden ExisNet, Inc
804-552-1009