Re: (PM) PM3 disconnect problems

Josh Richards (jrichard@livingston.com)
Sat, 9 May 1998 00:45:36 -0700 (PDT)

On 8 May 1998, Tim Wilfong wrote:

> I've seen other ISPs on this list complaining about seemingly random, but
> frequent disconnects for users dialling into their PM3s. It seems that
> either this is a relatively rare occurance, or there is no simple answer
> to how to fix it.

And there are many reasons and things that can cause this too,
unfortunately. Modem are imperfect things, running over imperfect
telephone networks... :)

> I've had problems with a small, but significant portion of our users
> dialing in and getting disconnected in the midle of transfers (no, not
> idle disconnects or anything obvious like that) -- the disconnects usually
> happen anywhere between 1 minute and 20 minutes after logging in. The
> reported cause is always User-Request, even though it is never a user
> request.

Any particular brand of modem or chipset? What types of connection speeds
are these users getting (look at a "show mX" then find the sX port they
are connected to)? How are your lines provisioned? A User-Request means
that the user was either reset from the CLI (or Amber), or the PM received
a terminate request (which is how most PPP client stacks disconnect). Turn
on a "set debug 0x51" watch it for a problem user until they get
disconnected and see if anything goes across (from either side), then run
it through the PPP decoder on our web site.

> We're currently running 3.7.2, but we've had these problems with earlier
> versions as well (although not as much).

You should be using 3.7.2c3 which is the latest modem code (well besides
a beta release).

-jr

----
Josh Richards - <jrichard@livingston.com> - [Beta Engineer]
LUCENT Technologies - Remote Access Business Unit
(formerly Livingston Enterprises, Inc.)
http://www.livingston.com/

-
To unsubscribe, email 'majordomo@livingston.com' with
'unsubscribe portmaster-users' in the body of the message.
Searchable list archive: <URL:http://www.livingston.com/Tech/archive/>