Fedora Networking :: VLAN Do Not Work Over Bonded Or Bridged Interfaces
Aug 20, 2011
I'm having trouble getting Fedora 15 to work with my em/bond/br configuration as it did in Fedora 14. I've got a bonding interface over em1 and em2, and then individual bond vlans (bond0.2, bond0.10, etc...), I then have a similarly named bridge interface (br0.2, br0.10, etc...) for kvm. Regardless of if I assign bond0.2 or br0.2 the IP address, I am unable to ping the gateway, but the native untagged VLAN (bond0, br0) has no issue at all.
To try to isolate this issue down to the switch or Fedora 15, I went with a plain em setup and configured em1, and em1.2 which works as expected. This issue has been plaguing me since Fedora 15 came out and I'm about to trash it and go back to Fedora 14 (Which didn't have this issue), but some of the new kvm features have kept me trying to get over this incomprehensible hump.
I have a rack of IBM x335s. They are all running 802.1Q over 802.3ad, that is VLAN trunking over two physical Ethernet bonded with LACP. Its been working fine on all of the servers for years.One server recently had its hard drives upgraded and CentOS reinstalled. It is a minimal install. The server has been reconfigured exactly as it was and has been verified and compared against the other working servers.It will not enslave the interfaces on boot. You can manually enslave and then bring up the network, so that much of the configuration is verified. It simply refuses to enslave... which fails everything that comes after thatHere is a snippet of the message log on the machine that does not enslave on boot
Jan 31 22:28:57 vm3srvp01 kernel: NET: Registered protocol family 10 Jan 31 22:28:57 vm3srvp01 kernel: lo: Disabled Privacy Extensions Jan 31 22:28:57 vm3srvp01 kernel: IPv6 over IPv4 tunneling driver
Just something that struck me while working on our virtual servers today.
I have bonded 3 NICs at the host in Ubuntu Server 8.04 LTS. They are using mode 0 for Round-robin. Point is to increase the speed/performance of all the servers, but mainly the fileserver. The fileserver is a virtual server running Ubuntu Server 8.04 LTS on VMware Server 2.0.
1) I noticed the NIC in the slave OS reported link speed as 1000 and Im unable to change it as the NIC (virtual one) doesnt support it. Does this not really matter, as the NIC doesnt exist, and it will run at higher speeds anyway? Or do I have to remove the bond on the host, bridge all 3 interfaces from the host to the slave OS, and then make a bond in the slave OS?
2) While at it, does mode 0 only increase performance on data being sent from the host or does it also increase the available incoming bandwidth?
I have two eth interfaces bridged in CentOS, one of these interface is connected to a Windows Server, I can do a ifconfig down on both these interfaces and there is no change of status at the windows side, I don't get the message that 'network cable has been unplugged'
Is this behavior normal ? I think when you shut an interface than you get the interface down message at other end ? And what becomes the status of the bridge if I shutdown one of its interface ?
I'm searching for a small perl script to iterate through the configured eth and vlan interfaces to get their assigned ip address and perform some processing.I'm after something like:
@interfaces = (waytogetinterfaces); foreach (@interfaces){ #perform processing print "Interface IP is: ", $_->IP; }
Using Fedora 10, can anyone tell me how to setup the network scripts to create two network interfaces for vlan x and y. Both interfaces should obtain an ip from dhcp and both interfaces should run over eth0.
Can anyone tell me how to setup the network scripts to create two network interfaces for vlan x and y. Both interfaces should obtain an ip from dhcp and both interfaces should run over eth0.
- eth0 is connected to a 'Green" interface of a smoothwall router. - eth1 is connected to a 'Orange' interface of the same smoothwall router.
The smoothwall router is setup to forward port 80 to the address of the eth1. I can ping eth1 from the smoothwall router so there is that connection.I have a web server running on port 80 and I have opened port 80 in the firewall and made eth1 a trusted interface.In this configuration, when accessing the outside world all the traffic goes through eth0 and DNS is snappy and faster. The problem is that anyone outside can not access the web server with a timeout error.If I disable eth0 and set DNS info on eth1( DNS servers are the QWest IPs ), people can access the web server, but DNS is very poor often taking more than a minute for a look up.Is there any way to get the two interfaces to work together?
I have 3 Interfaces for a different LAN's and when I start one interface the another interfaces goes down.How can it's possible?I configure my ethernets as:
Im using fedora 13 x64 install from DVD. Has anyone else found problems using vmware 7.0.1 or 7.1. My bridged networking will not pickup a dhcp address, and even if I configure static my network will still not work, yet NAT works fine. Even configuring bridged to use a dedicated device doesnt work. Workstation worked fine on Fedora 12 but this seems to be a reoccuring problem in every other release.
Bridged successfully but cannot access the Internet.I installed VMware in my Fc13,and installed XP sp3 in the VMware.I can access the Internet in XP,I set the network as Bridged with Fc13,but Fc cannot open a page,but I ping google.com,it can display: Code: Pinging google.com [64.233.183.104] with 32 bytes of data: then nothing.I still cannot access the Internet in my Fc13.
I have been trying to set up bridged networking, but I keep failing. I am using Fedora 14 x86_64 KDE as host with qemu-kvm and SPICE. The plan is to install a windows server, a few windows clients and then rawhide as guests on that. Naturally I want to use bridged networking for the windows guests.[URL]..But those both leave the guest without internet access. Is it really this difficult, or am I doing it wrong(tm)?
I'm preparing to "follow" these directions which will make VirtualBox's networking behave like VMWare's "NAT" networking. Of course, these instructions are for a Ubuntu/Debian setup.
I am quite certain that I can write a script that'll do the equivalent steps on my Fedora 11 box -- which is acceptable to me. I would, however, like to know how to splice this stuff into NetworkManager and/or /etc/sysconfig/networking/devices/, to do this the "right" way.
The desired networking behavior: -VM's get assigned addresses via DHCP on a virtual network -VM's can talk to each other and to their physical host -VM traffic out to the 'real world' is NATted
I just have installed FC 10 on a box that I will use as a router/firewall box. On the box I have three interfaces (eth0, eth0.704 (VLAN), and eth1). When the machine boots up, only eth0 and eth1 come up. If I run /etc/init.d/network restart, then eth0.704 will come up. If I add that command to rc.local, then all interfaces come up at boot.
I have the directive ONBOOT=yes in ifcfg-eth0.704. What would cause this vlan interface not to start at boot on the machine?
I've got a Fedora 13 box running a bunch of VM's w/ bridged networking.
For some reason, every 10-15 minutes the vm's networking freezes and the vm's become unresponsive to pings and any other network traffic for 10-15 seconds. It eventually gets restored.
I don't see any error messages other than this occasionally on the host: eth0: received packet with own address as source address
EDIT: I found a sample configuration, sadly without any comments, where there are defined some interfaces named vlan1:0 trough vlan1:3. Could that be the option I've been looking for.However in the definition of the vlan1 device it is configured with the subnetmask 255.0.0.0 . I would like to know the mechanics of that option, too.
2. In the wlan part of the network I want to have a dhcp server distributing ip addresses. However I want hosts the server 'knows' (by mac address) to be in a different vlan than those unknown. Is there any way to achieve that? If yes, do I need special wlan access points or can I do it by my configurable switch or by the server?
I have a weird issue that I have not seen on any forum. My jaunty on DELL studio laptop seems connected to net, but I can not access any network service (ssh, firefox etc.). But when I connect a cable the cable lights blink as it should be and in wireless connection my wifi light blinks.
It was working 2 days ago without problem, and I have not done big changes recently.I removed and reinstalled network-manager and network-manager-gnome. Nothing changed. I see a message in each restart as follows (when Openafs is starting). I can reproduce it with "/etc/init.d/openafs-client restart"
Code:
ADVISEADDR:error in specifying interfaces: no existing ip interfaces found
So I keep reading that bonding ethernet devices is supposed to be easy. I have followed several tutorials and tried a few things on my own, and I can't seem to get it right. I currently have a setup that allows me to ping internally on my network, but when I try to ping externally I get the error: connect: Network is unreachable
Before I bonded the interface my connection worked fine. Here are my config files. I am running centos 5.3 on a virtual machine. /etc/sysconfig/network-scripts/ifcfg-bond0
If I try to add a new interface (eth1) to /etc/network/interfaces, I get
Code: * Reconfiguring network interfaces... SIOCSIFADDR: No such device eth1: ERROR while getting interface flags: No such device SIOCSIFNETMASK: No such device
[Code]...
How do I add 2 interfaces and get anyone of them to work, as available ?
I have installed qemu/kvm and created a Bridged network connection which works just fine(Windows 7 VM won't work in NAT mode.)
But when I try to use NetworkManager it says that I have no network connection because the network isn't managed, (I set the settings in ifcfg-br0 and ifcfg-eth0 to be managed)
The real problem is that now I can't use my VPN connections (I have many) in NetworkManager.
Is there a way to have both of these pieces of functionality?
I am running Fedora 13 and after my machine is booted my ethernet interface eth0 does not have an IP address. Running
Code:
At the terminal does not work, and I can see that in /var/log/messages I have the following error:
Code:
I assume that the same command was attempted during the boot process (please correct me if I am wrong) to use DHCP to get an address for eth0, but failed for the same reason it is failing after boot when I run it manually, whatever that reason may be.
I noted, however, that eth0 IS in broadcast mode:
Code:
I also noted that running
Code:
Does in fact work and will run DHCP and configure eth0 with an ip address.
It is quite mysterious to me why running 'dhclient eth0' will work, but manually running 'dhclient' does not.
My brother has a Ubuntu server attached to a LAN with Windows PCs. I set up an openVPN tun service on the server, let's call this VPN1, so that I can connect remotely from my Ubuntu desktop. The server has one NIC and the LAN has a router that is the gateway to the internet.
My brother would like to remotely access his windows network when roaming with his Windows laptop.
Therefore, I would like to set up a second VPN service, this time tap, that is bridged with his LAN. Let's call this VPN2.
The LAN subnet is 192.168.1.0/24. The internet gateway is 192.168.1.1
The NIC has a fixed IP set by the router of 192.168.1.150
How do I set up VPN2 so as not to trash VPN1? That is, if I bridge eth0 with tap0 how will that impact VPN1's tun0 that is using eth0? Do I need a separate NIC for the VPN2 bridge?
I am trying to get a bridged connection to work in centos
Here is the network diagram [url]
There are two interfaces in linux system , they are bridged and connected to the windows system I am not sure if I need to enable STP in the bridge or not?
I found problem in FC 12 release. I installed fc 12 to server (earlier there was FC 7). Then as always i realised that names of interfaces (eth0, eth1, eth2 .... eth9) changed after install. eth0 became eth9, eth1 became eth5 ... Earlier on previous fedora releases i solved this problem correcting HWADDR in ifcfg-eth files, so i linked MAC addresses to names (eth0, eth1, eth2) as i wanted in a right order. Now I can't do this. After correcting these files and restarting network sevice i constantly get a message: device ... has different mac than expected. I looked ifup-eth file and compare it with the same one in fedora 7. And i found that in fedora 7 there is function rename_device which processes if HWADDR value doen't coincide with real mac address value. See code:
# remap, if the device is bound with a MAC address and not the right device num # bail out, if the MAC does not fit if [ -n "${HWADDR}" ]; then FOUNDMACADDR=`get_hwaddr ${REALDEVICE}` if [ "${FOUNDMACADDR}" != "${HWADDR}" ]; then curdev=`get_device_by_hwaddr ${HWADDR}` if [ -n "$curdev" ]; then rename_device "${REALDEVICE}" "${HWADDR}" "${curdev}" || { echo $"Device ${DEVICE} has different MAC address than expected, ignoring." [Code].....
But in fc 12 release there is no function rename_device!!!!!!!!! So in this case if HWADDR value is not the same as FOUNDMACADDR value (which equal to REALDEVICE mac address) i just get an error message So I can't change interfaces names, as result i can't organize right order of network interfaces as it was earlier on fc7.
When ever I restart the machine, both interfaces pull a IP, but the it randomly chooses eth0 or eth1 to have the gateway. I want the gateway to be eth0 always.
[Code]...
Is there some file or setting I can modify in Fedora10 that will always choose 192.168.1.1 on eth0 as the default gateway?
