Hi Floris,
On Feb 6, 2011, at 00:11 , Floris Bos wrote:
> I'm having issues getting SL6 installed on systems with more than one network
> interface.
> Since the network interface could be either eth0 or eth1 depending on the
> server, we normally specify the kickstart network device by MAC-address as
> boot option, e.g.:
>
> ==
> ksdevice=11:22:33:44:55:66
> ==
>
> And specify a static IP-address (the same as assigned by DHCP), without
> explicity naming an Ethernet device in the kickstart file, e.g.:
>
> ==
> network --bootproto=static --ip=10.0.0.10 --netmask=255.255.255.0 --
> gateway=10.0.0.1 --nameserver=127.0.0.1 --hostname=hostname
> ==
>
> Under Centos 5.5 and Fedora this works correctly, and causes the network
> device specified by ksdevice to be used.
>
> However it does not seems to work with SL6 beta.
this is probably a "6 vs. 5" issue rather than am "SL vs. CentOS/Fedora" one.
> It does get a DHCP lease on the right interface, downloads the kickstart file
> and the first installer files, but then it always configures "eth0" with the
> network information specified, instead of the actual ksdevice.
>
> This is an issue if the ksdevice is actually eth1.
There are significant changes in the handling of the interfaces from EL5 to EL6. It's quite possible that "ksdevice=<mac address>" no longer works.
> Because eth1 was assigned the IP-address by DHCP, and SL assigns the same IP-
> address to eth0 as well, network connectivity breaks, and it throws an error
> that it is unable to download the other installation packages.
Unless you have a special reason to have the interface named eth1, it may be better to have it recognized as eth0 in the first place. We're now generally using "ksdevice=link pci=bfsort" when kickstarting, which makes the "primary" NIC (as seen by the BIOS/vendor) eth0 during installation on all our servers which would otherwise recognize it as eth1, avoiding all these problems.
Best regards,
Stephan
--
Stephan Wiesand
DESY -DV-
Platanenenallee 6
15738 Zeuthen, Germany
|
