Another possibility (and better news if this one's right) is that it's a
ComOS bug. There was apparently a hard-to-find bug in 3.1.3c2. It worked
for just about everyone else, but all three pm's on the LAN at my location
started cutting people off mid-session. SOMETHING on our LAN tripped this
bug. This would happen to ANY USER who was PAP-authenticated, without
fail. We downgraded back to 3.1.2 (with filters of course) for a while and
the problem disappeared. I'm happy to say 3.1.4 seems to have squashed
this bug as well. All of our pm's now run 3.1.4. The tell-tale sign of
this bug is a disconnect followed within one second by a login-failed
(after the user has already been logged in for some time!). Here are the
five pertinent lines from termlog:
Sep 03 12:03:14 pm3 dialnet: port S10 PPP succeeded dest Negotiated
Sep 03 12:03:18 pm3 dialnet: port S10 joeblow succeeded dest 204.57.120.231
Sep 03 12:03:18 pm3 dialnet: port S10 connection succeeded dest link231.txdirect.net
...time passes...irrelevant stuff removed...
Sep 03 13:42:50 pm3 dialnet: port S10 session disconnected dest link231.txdirect.net
Sep 03 13:42:50 pm3 dialnet: port S10 joeblow login failed
Notice that the last two lines occur at the same time, but the "login
failed" is always after the disconnect.
Hope that's the prob...easy fix! Upgrade to 3.1.4.
Michael Nerone | Internet Direct, Inc. | http://www.txdirect.net
nerone@txdirect.net | 722-B Isom Rd. | Please direct all queries
| San Antonio, TX 78216 | to sales@txdirect.net.
| Voice: (210)308-9800 | Direct all tech questions
| Fax: (210)308-9240 | to support@txdirect.net
Any opinions expressed herein are my own and do
not necessarily reflect those of my employer.
On Fri, 22 Sep 1995, John R. Jamerson wrote:
> I originally bunged most of this onto the "DTR Dropping" thread, then thought
> better of it. Sorry if it's redundant to some of you.
>
> Let me preface this by saying that all ports idletime are set to 10 mins, except
> my backdoor port which is set to zero (no time out). We're an ISP, using
> PM's for remote POP's. We've got 10 PM's scattered all over.
> The problems are not site specific, but network wide.
>
> Got an ongoing beef from some users that they get kicked off the system
> after a minute or two with a "NO CARRIER" msg. Other users say the duration
> of the session prior to kick off varies. Got other users who either aren't
> getting kicked
> off, or are to lazy/whatever to gripe.
>
> I personally get kicked off the system after between one and two minutes _every
> time_ I call in from home -- and trust me, it's not an inactivity problem in
> my case.
> Then when I call right back, I can be on indefinitely.
>
> (This is making me crazy, can you tell???)
>
> And this situation happens no matter whether I dial into the backdoor port or
> the common modem bank. I live smack between two of our sites, so both are
> local calls from the house. It happens on either/both PM's. I can also dial
> long distance to verify this from our other PM's.
>
> The user's software and hardware (modems) vary across the spectrum.
> Text users have Procomm (and +), Telix, Windows Terminal (shudder), Hayes
> Smartcomm, etc. SLIP and PPP users use Trumpet Winsock, MacPPP, OS/2
> whatever, Win 95, you name it. Modems -- I've been kicked off using Microcom,
> Hayes, USR, Zoom, PPP -- everything from an old Hayes Personal 2400 thru
> USR and PPP 28.8's.
>
> On the POP end, we're using Zoom's, Hayes, and Supras -- all sites are
> homogenous,
> no mixes at any site, except my backdoor modem's a 28.8 USR.
>
> We're using 3.1.2 PM software, SunOS 4.1.3 on the server and using Radius
> for authentication.
>
> Maybe I can't see the forest, but I can't figure out what's causing this.
>
> Any further suggestions/comments/opinions welcome. And knowing this
> group, there'll be opinions, at least :-)
>
> TIA,
> John Jamerson
> Systems Administrator/WAN Manager
> CoastalNet/Global Information Exchange
> http://www.coastalnet.com
>
>