Christian Schmid
2005-May-06 10:22 UTC
Multipath Routing in same subnet - Please take a look
Hello.
I have the problem that I have two gateways on the same subnet, 80.237.244.1 and
80.237.244.33. Both
gateways are 100 MBit cards, so I have 2 times 100 MBit to the Internet. The NIC
in the server is a
gigabit-card, so this card is easy able to use both gateways for outgoing
traffic.
Now I just want to use both gateways for my outgoing traffic but no matter what
I do, it doesnt
work. I tried "ip route replace default scope global nexthop via
80.237.244.1 dev eth1 weight 100
nexthop via 80.237.244.33 dev eth1 weight 100" but it only either sends all
traffic to .1 or all to
.33. I have many downloaders so it shouldnt be a problem, should it? Its a
2.6.12rc3 kernel with
advanced router enabled and all multipath-options enabled as well.
"ip" doesnt give an error and "ip
route ls" shows the following:
80.237.244.0/26 dev eth1 proto kernel scope link src 80.237.244.52
default
nexthop via 80.237.244.1 dev eth1 weight 100
nexthop via 80.237.244.33 dev eth1 weight 100
I have read postings on the net but all of them are using huge scripts because
they are on different
networks. My problem seems to be a much easier problem but I just cant get this
to work..... :(
Please help.
Best regards,
Chris
Christian Schmid wrote:> > Hello.> 80.237.244.0/26 dev eth1 proto kernel scope link src 80.237.244.52 > default > nexthop via 80.237.244.1 dev eth1 weight 100 > nexthop via 80.237.244.33 dev eth1 weight 100Do not use weight parameters exceeding a single digit!> I have read postings on the net but all of them are using huge scripts because they are on different > networks. My problem seems to be a much easier problem but I just cant get this to work..... :( > > Please help. > > Best regards, > ChrisI''m no expert, but my suggestion is to use 2 NICs and connect one to each uplink or at least add 2 entries into /etc/iproute2/rt_tables. I think you''ll find a similar situation answered in the ML within the last 10 days or so, but I can''t recall the subject of the thread. Nor can I find anything specifying exactly what rt_tables needs to contain :/ You can review what I''ve gleaned from this ML at http://yesican.chsoft.biz/lartc The most urgent things for you to know: 1) The LARTC HOWTO is wrong 2) You must apply Julian''s patch 3) You _really_ need to read nano.txt 4) All of the intelligible success stories WRT multipath are either on yesican or are linked to from there. --gypsy
Alexander Samad
2005-May-07 04:25 UTC
Re: Multipath Routing in same subnet - Please take a look
On Fri, May 06, 2005 at 08:41:55PM -0700, gypsy wrote:> Christian Schmid wrote: > > > > Hello. > > > 80.237.244.0/26 dev eth1 proto kernel scope link src 80.237.244.52 > > default > > nexthop via 80.237.244.1 dev eth1 weight 100 > > nexthop via 80.237.244.33 dev eth1 weight 100 > > Do not use weight parameters exceeding a single digit! > > > I have read postings on the net but all of them are using huge scripts because they are on different > > networks. My problem seems to be a much easier problem but I just cant get this to work..... :( > > > > Please help. > > > > Best regards, > > Chris > > I''m no expert, but my suggestion is to use 2 NICs and connect one to > each uplink or at least add 2 entries into /etc/iproute2/rt_tables. I > think you''ll find a similar situation answered in the ML within the last > 10 days or so, but I can''t recall the subject of the thread. Nor can I > find anything specifying exactly what rt_tables needs to contain :/ > > You can review what I''ve gleaned from this ML at > http://yesican.chsoft.biz/lartc > > The most urgent things for you to know: > 1) The LARTC HOWTO is wrong > 2) You must apply Julian''s patch > 3) You _really_ need to read nano.txt > 4) All of the intelligible success stories WRT multipath are either on > yesican or are linked to from there.here is my setup (firewall with 2 default routes) eth0 = cable, ppp0 adsl. I changed the dhcpd client to add the default route for each in on a different metric, that way if the one of the lines is out the other default route will still work! from ip r (whith some stuff removed - but the essential stuff is here) default metric 5 nexthop via 141.168.16.1 dev eth0 weight 4 nexthop via 202.7.162.89 dev ppp0 weight 2 default via 141.168.16.1 dev eth0 metric 10 default via 202.7.162.89 dev ppp0 metric 20 from ip ru 0: from all lookup local 200: from 141.168.16.16 lookup cable 201: from 60.240.81.237 lookup adsl 32766: from all lookup main 32767: from all lookup default the important ones are 200, 201 these setup up the routing for each of the different legs - cause they will be different! so 141.168.16.16 is on eth0 and 60.240.81.237 is on ppp0 (they are the actual address on the interfaces) cat /etc/iproute2/rt_tables (these are where the names come from above!) # # reserved values # 255 local 254 main 253 default 0 unspec # # local # #1 inr.ruhep 200 cable 201 adsl These are the routing tables for 200 and 201, the 192.168 address are the local address. The routing engine workes from the lowest numbered ru (from ip ru ) and works to the larger numbered ones until it finds a rule that matches! # ip r sh tab 200 192.168.11.0/24 dev br0 scope link 192.168.10.0/24 dev eth3 scope link 192.168.9.0/24 dev eth4 scope link default via 141.168.16.1 dev eth0 default via 141.168.16.1 dev eth0 metric 10 # ip r sh tab 201 192.168.11.0/24 dev br0 scope link 192.168.10.0/24 dev eth3 scope link 192.168.9.0/24 dev eth4 scope link default via 202.7.162.89 dev ppp0 Hope that helps> > --gypsy > _______________________________________________ > LARTC mailing list > LARTC@mailman.ds9a.nl > http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc >_______________________________________________ LARTC mailing list LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
Andy Furniss
2005-May-09 23:40 UTC
Re: Multipath Routing in same subnet - Please take a look
Christian Schmid wrote:> nexthop via 80.237.244.1 dev eth1 weight 100 > nexthop via 80.237.244.33 dev eth1 weight 100 > > > I have read postings on the net but all of them are using huge scripts > because they are on different networks. My problem seems to be a much > easier problem but I just cant get this to work..... :(Saw this on netdev, solution disable IP_ROUTE_MULTIPATH_CACHED http://news.gmane.org/find-root.php?group=gmane.linux.network&article=25774 Andy.