CONFIG_IP_ROUTE_MULTIPATH=y CONFIG_IP_ROUTE_MULTIPATH_CACHED=y CONFIG_IP_ROUTE_MULTIPATH_RR=y CONFIG_IP_ROUTE_MULTIPATH_RANDOM=y CONFIG_IP_ROUTE_MULTIPATH_WRANDOM=y CONFIG_IP_ROUTE_MULTIPATH_DRR=y and all of this algorithms works perfectly over icmp , but there is some (maybe) problems with tcp round robin algorithm : with : ip route add default mpath rr nexthop via 1.1.1.2 dev eth0 nexthop via 1.1.1.3 dev eth0 wget i attempt to wget a file over tcp port 80 , all packets goes to 1.1.1.2 , nothing on 1.1.1.3 . but with : ip route add default mpath random nexthop via 1.1.1.2 dev eth0 nexthop via 1.1.1.3 dev eth0 i've got something on 1.1.1.3 and everything goes as expected , any idea ? we have to report bug ?
There are some problems with the balancing code in newer kernels. As stated few days earlier removing multipath caching does help prety good. Also a side remark with one client (load) you will see almost no effect. When you load the link wih lots of connections then the balancing works. Remove multipath caching and try again If your ISP's can offer you BGP (even read only/listen) resort to it. It will do much better. -- С уважение, Владимир Витков http://www.netsecad.com http://www.supportbg.com
_______________________________________________ LARTC mailing list LARTC@xxxxxxxxxxxxxxx http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
