Milter timeout message

Vernon Schryver vjs@calcite.rhyolite.com
Wed Jan 22 19:55:34 UTC 2003


> From: Sean Rima <sean@tcob1.net>

> >     Xdcc, S=unix:/var/run/dcc/dccm, F=T, T=C:30s;S:5s;R:5s;E:30s

> Is there a similar method for dccproc

The timeouts for sendmail waiting for a response from the local DCC
client are a different sort of thing than the timeouts of a DCC client
waiting for a DCC server.  Sendmail's milter timeouts to a
(usually) local milter are fixed, and probably should be.  The DCC
client timeouts are adaptive and depend on the average RTT to the distant
DCC server beyond the Internet horizon.

When things are working, your DCC client should not need to wait more
than 0.5 seconds.  Only when a local PPP or other slow speed link is
stalled by congestion or modem retraining are multi-second delays
reasonable.  Recent versions of the DCC client code can wait several
seconds for before giving up, assuming the previously preferred DCC
server has died, and trying with a new IP address.  (There are currently
8 IP addresses for `dcc.dcc-servers.net.)  That process might make
5 seconds too short.


>                                       as I sometimes get:
>
> Jan 22 18:17:08 tcob1 dccproc[29273]: no answer from
> dcc.dcc-servers.net (203.147.165.193,6277) after 12015 ms
> Jan 22 18:17:28 tcob1 dccproc[29308]: no answer from
> dcc.dcc-servers.net (203.147.165.193,6277) after 12024 ms
> Jan 22 18:17:41 tcob1 dccproc[29320]: no answer from
> dcc.dcc-servers.net (203.147.165.193,6277) after 12011 ms

If no DCC server has answered your DCC client after 12 seconds, and in
this particular case, apparently after 3 consecutive 12 second waits,
then waiting even longer is like to be a waste of time.

Isn't this problem you see the same that you've had for a long time and
that I've speculated is an IP routing or PPP interface configuration
problem?


Vernon Schryver    vjs@rhyolite.com



More information about the DCC mailing list

Contact vjs@rhyolite.com by mail or use the form.