Dracut with br0 causes network failure

[gdhgdh]

Hi,

Today I've been having fun trying to boot to an NFS root on CentOS 7 whilst setting up a bridge br0 device. It all works fine until I get the root: login prompt.. at which point the machine drops off the network. (pings from other network hosts to 10.0.0.62 stop..)

If I remove the br0 references from the cmdline in pxelinux.cfg/default then the machine boots correctly and all's good:

Code: Select all

label userver
  kernel userver/vmlinuz-3.10.0-123.20.1.el7.x86_64
  append initrd=userver/initramfs-3.10.0-123.20.1.el7.x86_64.img rw ip=10.0.0.62::10.0.0.254:255.255.255.0:userver2:enp3s0:off nameserver=10.0.0.254 root=nfs:10.0.0.22:/mnt/internal/boot/userver2 rd.debug rd.udev.debug fbcon=scrollback:1024k

However, I need that br0 bridge since the use case is diskless compute nodes for an OpenNebula cluster. This pxelinux.cfg/default entry fails:

Code: Select all

label userverbr0
  kernel userver/vmlinuz-3.10.0-123.20.1.el7.x86_64
  append initrd=userver/initramfs-3.10.0-123.20.1.el7.x86_64.img rw ip=10.0.0.62::10.0.0.254:255.255.255.0:userver2:br0:off bridge=br0:enp3s0 nameserver=10.0.0.254 root=nfs:10.0.0.22:/mnt/internal/boot/userver2 rd.debug rd.udev.debug fbcon=scrollback:1024k 

dracut does write out the correct ifcfg-br0 and ifcfg-enp3s0 to /etc/sysconfig/network-scripts and like I say, the getty login: prompt appears - immediately after that, all pings from other hosts to this one stop. :(

The output from the 'debug' mode of dracut looks good - I can see the relevant 'brctl br0 addbr' and 'brctl br0 addif enp3s0' commands being issued.. it just all goes south after the boot has finished, and I don't know how to get more information at that time, since the NFS root is lost and the machine becomes useless.

Any suggestions warmly welcomed :)

Cheers,
Gavin.