openssh unix dev - Nov 2001 - Connections suddenly closing

Hi, I have been having a problem with OpenSSH which afaict has _never_
happened prior to upgrading to 3.0.  (I always run the current release,
so I'd been on 2.9.9[p2] prior to 3.0.)

At a seemingly random time, I will lose my connection to the remote
host, for example:

Read from remote host crate.alongtheway.com: Connection reset by peer

Granted, I am not sure that that was the error _every_ time because
prior to noticing this trend I had been using xterms with "-e ssh"
options so the xterm would close before I got to see the error.  But a
couple days ago I turned off that switch so I could see what was
happening, and today it finally happened again and gave me the above
error.

Note that I searched for anything similar in the archives of this list
and found the following results:

http://marc.theaimsgroup.com/?l=openssh-unix-dev&w=2&r=1&s=Connection+reset+by+peer&q=b

However none of those seems to apply here, as in at least one case, the
connection to the remote host was _not idle._  A couple example cases:

1) Had two connections open to a FreeBSD machine at a remote location;
one was an idle shell, the other was actively running the 'tin'
newsreader.  All of a sudden the connection running the tin session
dumped on me, meanwhile the other connection remained open and fine.

2) A connection I had open to an OpenBSD 2.9 system on my home network
(through an internal firewall doing NAT) suddenly dropped also, it was
idle overnight, probably for around 6 or 7 hours.

3) The connection drop that occurred today was to yet another FreeBSD
system, and had only been idle for around an hour.


Constants with every case:

1) Client system is Linux (debian potato) running OpenSSH 3.0p1.

2) Server system is OpenSSH 3.0 (p1 in the FreeBSD case).

3) Connection is NATted.

Wish I had more info.. any suggestions?  Anyone else have ssh
connections mysteriously die?

-- 

Jim B.
vader at conflict.net

On Thu, Nov 15, 2001 at 05:52:30PM +0000, Jim Breton wrote:
> Wish I had more info.. any suggestions?  Anyone else have ssh
> connections mysteriously die?
i occasionally see connections fail to establish, although they don;t seem to
die after that.  my server is OpenBSD 2.6, client is Redhat 7.2.  Both boxes
have the original versions of ssh installed as shipped.

my server runs a CVS repository, which is accessed via ssh.  a cron job on the
linux client does a cvs update every minute.  about 6-8 times per 24 hour
period,
cron sends me this email:

  Connection closed by 192.168.2.1
  cvs [update aborted]: end of file from server (consult above messages if any)

on the server side, i usually get three lines in authlog.  when it fails, i
only get the first.

  Nov 15 02:29:01 server sshd[32358]: log: Connection from client.my.com port
46460
  Nov 15 02:29:01 server sshd[32358]: log: RSA authentication for user accepted.
  Nov 15 02:29:02 server sshd[32358]: log: Closing connection to 192.168.2.2

  Nov 15 02:30:02 server sshd[11839]: log: Connection from client.my.com port
46465

  Nov 15 02:31:01 server sshd[11977]: log: Connection from client.my.com port
46470
  Nov 15 02:31:02 server sshd[11977]: log: RSA authentication for user accepted.
  Nov 15 02:31:03 server sshd[11977]: log: Closing connection to 192.168.2.2


i have been assuming that my upcoming upgrade to OpenBSD 2.9/3.0 will solve
this problem.  however, i'd be happy to enable more logging and dig deeper
if it might provide useful info for someone.


cheers

-- 
---------------------------------------------------------------------------
    Ray Lehtiniemi <rayl at mail.com> <rayl at otii.com>

some nat-gateways have small timeouts.

this is probably the cause for the problem.

is there a period of inactivity? or does
it happen while you type?