Centos 6 CT defaults to venet at boot

[jjsimpson]

Here is my issue:

I have server with two bridges, one for network A (vmbr0) and the other for B(vmbr1). These are two separate networks built under two separate models.

The default gateway is set to vmbr0 for cluster management and what not.


Here is the main issue, whenever I create a CT for machines running on network B I have to use veth connections so that they will talk with the outside world properly. I can go in and configure the veth and get it working.

The main issue I am having is that whenever I reboot one of my containers that has a veth on network B, it defaults the GATEWAYDEV in /etc/sysconfig/network to venet, and then I have to console in (or setup a venet and ssh) and change it back to eth0.

Is there a way to stop this from happening?

 

[jjsimpson]

So here is what I have figured out, venets and veths do not play well together. They work and you can route traffic through them, but if you have one of each, something in the boot process sets the venet to always be the gateway device. The only way to make a veth the gateway device is to make it the only type of interface that you use.

Even telling the venet to not start at boot does not work, it will turn the venet on and keep it as the gateway. I tested by creating a new CT and only using a veth, then I entered the CT through the host by using vzctl enter xxx and then I configured eth0. The CT then keeps the veth as the gateway at boot (because it is the only device outside of the loopback).

So if you are using multiple networks on one proxmox machine, it looks like the only way to make containers work properly on the second "non-gateway" interface is to use veths.


Anyone else know of a better solution?