Hi, I've been trying to set up a full redundant web system using loadbalancing technics (ultramonkey). To incorporate a redundancy in the network itself (in case a switch fails) each machine behind the loadbalancer has two internal nics on two diffrent networks.
(external ip) eth0 loadbalancer eth1 eth2 192.168.7.2 192.168.8.2 ¦ ¦ ¦ ¦ ¦ ¦ 192.168.7.4 192.168.8.4 eth0 eth1 webserver1 default gateway for eth0 (webserver1) is 192.168.7.2 default gateway for eth1 (webserver1) is 192.168.8.2 My problem is the following: I set up load balancing which works fine when only one nic is up on the web server. I can ping -I eth0 www.google.com or ping -I eth1 www.google.com respectevely when eth0 or eth1 is up. The problems start when both cards are up. I can only ping with one interface outside of the local net (local pings works on both nics). It is always the last card ifup'ed that can ping. If I don't specify the interface while pinging it works fine using the last ifup'ed interface. My biggest problem is that the load balancer thinks both nics are up and that the service is available on both, thus every second time i get no response from the web server. I'm pretty shure this has to do with having two gateways for a single machine. When I remove the default route of the first card (the one that works) I can suddenly ping the outside world on the second card. I used tcpdump while pinging with the card that doesn't work. I can see the packages returing from the right source (if I ping -I eth0 64.233.179.104 which doesnt work and look at tcp dump on eth0 I see the returning packages with the right source and destination IP)... I have no idea why these packages are discarded... Any one an idea ? Thanks for any help Jean-Philippe -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]