Hi! In our students' hostel we have 6 DSL lines (dialups to different providers); we have set up a linux box (currently running 2.6.11-rc2-mm2, but the problem described hereafter also applies to previous 2.6-series kernels) with help from http://lartc.org/howto/lartc.rpdb.multiple-links.html#AEN298 Our Setup looks like this: 10.0.0.0/8 10.70.255.1 +----------+ +-----------+ | intranet |---| linux-box | +----------+ +-----------+ 10.254.0.1 | | | +-----------+ | | +-----| HW-router |---(DSL -> Provider) | | +-----------+ | | | +--------- (see above) [10.254.0.2] +------------- (see above) [10.254.0.3] (Above schematics drawn with only 3 instead of 6 DSL links) Our problem is as follows: it seems that when the route cache expires also existing TCP connections are rerouted, causing the connection to get lost. (Just a theory - I don't know how to check that) While that is fully comprehensible with UDP traffic, I thought I read somewhere that this shouldn't apply to TCP traffic (connections - once established - will always be routed through the initial hop). Does anyone know how to avoid this problem (which makes downloading big files a pain as some download-manager has to be used that supports appending/byte ranges)? Am I doing something wrong here, did I forget something? Our Setup is as simple as setting a multihop default route on the linux box like this: ip route add default proto static \ nexthop via 10.254.0.2 dev eth0 \ nexthop via 10.254.0.3 dev eth0 \ nexthop via 10.254.0.1 dev eth0 \ nexthop via 10.254.0.4 dev eth0 \ nexthop via 10.254.0.5 dev eth0 We don't do any NAT as this is done by the hardware DSL routers. Thank you for any help! Marcus ------------------------------------------------------------------------- The most difficult thing in the world is to know how to do a thing and to watch somebody else doing it wrong, without comment. _______________________________________________ LARTC mailing list / LARTC@xxxxxxxxxxxxxxx http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/