[Savannah-hackers] [support #103027] premature ssh timeouts on slow connections

[Sylvain Beucler]

This mail is an automated notification from the support tracker
 of the project: Savannah Administration.

/**************************************************************************/
[support #103027] Latest Modifications:

Changes by: 
                Sylvain Beucler <address@hidden>
'Date: 
                Thu 04/29/04 at 18:19 (Europe/Paris)

            What     | Removed                   | Added
---------------------------------------------------------------------------
          Resolution | None                      | Fixed
              Status | Open                      | Closed







/**************************************************************************/
[support #103027] Full Item Snapshot:

URL: <http://savannah.gnu.org/support/?func=detailitem&item_id=103027>
Project: Savannah Administration
Submitted by: Karl Berry
On: Fri 04/23/04 at 21:59

Category:  None
Priority:  5 - Normal
Severity:  3 - Ordinary
Resolution:  Fixed
Assigned to:  Beuc
Originator Email:  address@hidden
Status:  Closed


Summary:  premature ssh timeouts on slow connections

Original Submission:  I unfortunately had to move to a satellite-based Internet 
connection, since nothing else was available in my area.  It has very slow 
upload speeds with a large latency.  

Since moving, I've found that cvs updates in larger hierarchies, namely 
/webcvs/texinfo, usually die after approximately 90 seconds with the message:
Received disconnect from 199.232.41.3: 2: Timeout, your session not responding.

It will occasionally run to completion, occasionally time out sooner, etc.  
I've tried everything I can think of on the client side to affect this: 
ServerAliveInterval, ServerAliveCountMax, TCPKeepAlive, and more, but nothing 
has made any difference.

On the server side, setting ClientAliveInterval 15 (or some such) in 
sshd_config seems like the most likely thing to help, although I obviously 
don't know for sure.

Are you willing and able to try to help with this?  I don't know if anyone else 
is on a satellite connection; I doubt anyone on real broadband or even dialup 
would have the problem.

Thanks.



Follow-up Comments
------------------


-------------------------------------------------------
Date: Sun 04/25/04 at 13:42         By: karl
thanks very much! it seems to have helped, at least i was able to do a couple 
of cvs updates this morning without getting timed out.



-------------------------------------------------------
Date: Sun 04/25/04 at 08:41         By: Beuc
I set it to 20.
I am not sure that allowing a non-responsive client for 10 minutes will solve 
the problem, but it does not cost anything to try.
I will also see with the system administrators whether there is any cons in 
this change.


-------------------------------------------------------
Date: Sat 04/24/04 at 23:07         By: karl
Hi Sylvain,

The 15 was just a random example, ignore that.

Would you be willing to try increasing ClientAliveCountMax to something much 
larger, say, 10 or 20?  I don't think this would do any harm, as it will only 
affect clients that aren't successfully connecting anyway.  I don't know if it 
will help my case, but it's the only thing I can think of.

As for the "inactivity", it would take substantial debugging to figure out 
exactly why it is happening.  Since the timeouts usually occur only in 
hierarchies with relative large numbers of files, I surmise that the client has 
to send lots of data in quick succession, and the satellite connection simply 
won't pass it that quickly. With latencies around 800-900ms, anything that 
requires a round-trip handshake is not going to be completed quickly.

Thanks.


-------------------------------------------------------
Date: Sat 04/24/04 at 20:04         By: Beuc
Hello,

The current settings are ClientAliveInterval 30 and ClientAliveCountMax to 
default (3), hence the 90 seconds I guess :)

I do not understand your suggestion to set it to 15, as an unreachable client 
would be killed in 45 seconds.
I am also puzzled to see that your connection can be 'inactive' for more than 
one minute.
Is there something I missed?

Thanks.













For detailed info, follow this link:
<http://savannah.gnu.org/support/?func=detailitem&item_id=103027>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/