Adam Megacz wrote:
> I've encountered a NAT that seems to be messing with OpenAFS (1.4.4
> Mac client and 1.4.4 Linux server) -- periodically and randomly things
> just stop working and the connection times out; a few minutes later
> they magically start working again.
> 
> Are there any guides out there on how to debug such a situation?
> 
>   - a

The NAT is dropping the UDP port mapping.  So when a client tries to
send on an existing RX connection to a server, the server replies to
the original IP address / port number associated with the RX connection
and the message never reaches the client.  The client will timeout and
destroy the connection and mark the server as down.  When the next
server probe cycle occurs the ping will succeed and the server will be
marked as up.

There is a behavior modification we made to the Windows client that has
not made it into a release UNIX client where a request that ends in a
timeout is always retried once using a new RX connection.  We only mark
the server down if the request on the new RX connection fails.

Jeffrey Altman
Secure Endpoints Inc.

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature

Reply via email to