On Thu, Jan 29, 2009 at 2:23 AM, andrelst <[email protected]> wrote: > On Wed, Jan 28, 2009 at 3:23 AM, Ozzie de Leon <[email protected]> wrote: >> i am NOT smarter than a fifth grader! >> >> IP addresses are not bound to interfaces, but to the host [which is what >> everybody's been saying]. > > Ozzie, > > IP Address are bound to the NIC or logical NIC, and not the host.
andre.. that is not the case in linux.. you can read it here regarding to "Understanding Linux Network Internals" http://book.chinaunix.net/special/ebook/oreilly/Understanding_Linux_Network_Internals/0596002556/understandlni-CHP-28-SECT-4.html#understandlni-CHP-28-FN3 thats the reason why i still prefer xBSDs over linux because of the technical correctness done by the xBSDs... seen last week another uncovered bug on the TCP/IP stack of linux and i just made bandaid solution to circumvent this bug on our production servers using redhat advance server both for version 4.x and 5.x.. tsk tsk tsk >> so pinging the static [172.16.17.95] and the dynamic [172.16.17.15] was >> actually being routed through eth0 instead of >> eth1. with the cable connected to just eth1, only the static IP >> [172.16.17.95] could be pinged since eth0 couldn't connect to a DHCP server >> to pick up the dynamic address. > > Absolutely correct. And is expected behavior. However and take note, that > DHCP client will set one NIC, example is eth0 as the default > gateway(one default route), > not eth1. 2 multiple nics on one subnet is never a good idea in the first > place(google it and you'll see why.), hard to administer and hard to > troubleshoot later with the network or firewall guys. no not really... as long as they understand the communications between end to end on the same subnet... that is what ARP is all about... > To the list, let me know if you have done ip's on one subnet on two nics > without > bonding in dev or production.... i'm curious to know why. one example is apache's IP-based virtual hosting... to increase more network bandwidth on a given host without resulting to mutlitple subnets.. with the help of its networking tools.. you can redirect traffic that if the traffic from source IP A bond to device A send to device A.. if the traffic from source IP B bond to device B send it to device B.. and so on and so forth... thus you are having a multiple gateways based on the source IP address and not on the destination IP address.. > Just because you can does not mean you have to. That's why I mentioned about a > "simple solution" and sticking to the KISS principle... two ip's on one nic > and > consequently, on one default gateway. Less headache. well if that nic goes down.. you got a headache compare to multiple nics.. hehehehe although channel bonding, CARP and other high availability technology is out there for other solutions to his problem.. as the saying goes... to each his own style and right tool/technology for the right job... > Windows does what I'm saying... if it cannot reach a dhcp server, then > it sets a static > ip address of 169.xxx.xxx.xxx on the NIC. From the top of my head, I > think this > behavior is from an RFC I read few years ago. it is not static...it is still dynamic... RFC 3927 - dynamic configuation of ipv4 link local addresses.. it is intended for zero configuration networking... microsoft called this Automatic Private IP Addressing (APIPA).. fooler. _________________________________________________ Philippine Linux Users' Group (PLUG) Mailing List http://lists.linux.org.ph/mailman/listinfo/plug Searchable Archives: http://archives.free.net.ph
