vagab0nd - 2011-10-18

Passing along a band-aid to help with hardware based installs of NST 2.15.  I would be interested in knowing if other people experienced any similar problems and if this band-aid helped them.

Short story:
Despite the new FC15 feature of “consistent device naming”, add ifcfg-ethx entries for all your NICS ala the following; to avoid static IP addresses mysteriously getting duplicated on other interfaces:
DEVICE=ethx
BOOTPROTO=none
ONBOOT=no

The onboot parameter choice seems to be ignored, so a value of yes or no still results in an active NIC.

Long story
The problem is that,  at boot time on a HP Proliant, em1 ends up with the same IP address as em2. em1 is not listed with NetworkManger and /etc/network/interfaces does not exist. Em1 operates as expected with ifdown em1;ifup em1; but that does not provide a bootable solution.  In fact that the down/up to be done at the local console.

The band-aid noted above did work.

Measures that did NOT work for a bootable solution:

unique ifcfg-em1 in /etc/sysconfig/network-scripts (even with HWADDR!) per http://24.97.150.195/nstwiki/index.php/Hard_Disk_Install_Setup_Tips
DEVICE=em1
BOOTPROTO=none
ONBOOT=yes

ifdown and up of em1 in rc.local

ifcfg-em1 entry of onboot=no

running nstpostupdate after patching via yumex

ifdown and up of em1 in rc.last (not a convention in Fedora that I know of; but it is available in SUSE, so I gave it a try)

The ifcfg-em2 entry (slightly redacted) is:

DEVICE=em2
BOOTPROTO=static
IPADDR=aaa.bbb.ccc.102
NETMASK=255.255.255.0
NETWORK=aaa.bbb.ccc.0
BROADCAST=aaa.bbb.ccc.255
HWADDR=macaddress (this was included and excluded in testing)
GATEWAY=aaa.bbb.ccc.1
ONBOOT=yes
DNS1=aaa.bbb.ccc.242
DNS2=208.67.222.222
DOMAIN=this.that.something

What is ALSO very puzzling  is that while udev is renaming eth0 and eth1 per:
http://fedoraproject.org/wiki/Features/ConsistentNetworkDeviceNaming

…. apparently it does not always rename them the same in any given reboot per dmsg inspection of booting the HP Proliant; yet the result of em1 being the probe interface always ends up being correct.  I guess it is supposed to according to the previous URL, but it makes things pretty confusing.

V
'