Strange behavior of squid 2.0/2.1 tcp/udp_incoming/outgoing_address

Hi.

  I have written about it several days ago - but now some more things
'bout it.

  We have a local network (>250) and 2*115.2 (sl0&sl1) internet
connection. All www traffic is passed throught squid (on Linux 2.1.13x).
Our idea was to have all www trafic on the second modem (sl1). So we have
assigned an ip alias for our w3cache and forced squid to bind all traffic
to that alias-interface (eth0:1) (tcp_(incoming|outgoing)_address and
udp_outgoing_address options in squid's config file). We do not fetch data
from www hosts directly - cache is configured to use only parent cache
server) Routing alias-ip (sl0:1) was passed throught sl1 (2nd modem),
from other ip's in our network - throught sl0 (first modem)

  But the problem is that squid does not performs connection to parent
cahce using ip number that we told him to (ip from eth0:1), but also from
eth0. You know what that mean.

  I think that this situation takes place when one-modem data flow is not
enough to fulfill squid's bandwidth needs. In cache_log i can see 'TCP
connection to ......... failed and after about 30 seconds (...) succeeded'

  So i think that something is broken with
(tcp|udp)_(incoming|outgoing)_address. Or maybe i'm wrong and i've missed
sth.

  If You find it usefull i can send squid config file.

                                pozdrawiam
                  ___________ _____
         Maciej Szulc \ / e-mail:
  Student of Computer Science \/ wodzu@pg.gda.pl
Technical University of Gdansk /\
~~~~~~~~~~~~~~~~~~~~~~~~~~~~oooooo~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
         Student's hostel nr 2 network administrator
Received on Sat Dec 05 1998 - 17:28:03 MST