[OpenAFS] Connection timeouts

[ted creedon]

The windows client stays connected more reliably thru 2 NATTED firewalls
than the Linux client. If the Linux client were upgraded to do whatever the
windows client does the "cellname workaround" would be acceptable.

Connectionless UDP packets are port forwarded to the Class C server. I
recall previous mailings suggesting changing the timeout on the at the
firewalls.

That being said, is it possible for the server to reply with a keep alive
no-op?

tedc
-----Original Message-----
From: Derek Atkins [mailto:warlord@MIT.EDU] 
Sent: Tuesday, February 14, 2006 6:49 AM
To: tcreedon@easystreet.com
Cc: openafs-info@openafs.org
Subject: Re: [OpenAFS] Connection timeouts

"ted creedon" <tcreedon@easystreet.com> writes:

> For a client/server combination each behind a firewall:
>
>  1. Why does the Linux client timeout fairly rapidly requiring a client
>     restart?

NAT UDP Timeouts.  The firewall/NAT gateway loses the UDP mapping
between the client and the server.. The server can no longer talk
to the client..  Callbacks fail..  The server marks the client as
"Bad" because it can't talk back to the client.

-derek
-- 
       Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
       Member, MIT Student Information Processing Board  (SIPB)
       URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
       warlord@MIT.EDU                        PGP key available