On 7/22/2011 12:15 PM, Paul Heinlein wrote:
> On Fri, 22 Jul 2011, Les Mikesell wrote:
>
>>> I use ksdevice= as a boot option (e.g., in the APPEND section of
>>> syslinux or pxelinux config).
>>
>> How do you know which device is going to be eth0 at that point?
>
> That's a great question. Sadly, my answer is somewhat idiosyncratic
> and less likely to be of use in larger environments.
>
> In our small-office environment, I rarely kickstart a new server until
> after I've booted it into a rescue environment. The initial boot
> allows me to inventory the MACs (for dhcp), adjust IPMI network
> settings, check for driver issues, and double-check to ensure that the
> hardware matches the order. (I've only had one case where a vendor
> shipped less RAM than we'd purchased, but I'm glad I identified the
> problem before I wrote any data to the hard drives.)
I use ocsinventory-ng for this, but it doesn't report until everything
including the agent is installed and on the right network.
> It's during the MAC inventory that I identify eth0, eth1, ... I've
> never had a case where the order in the initial boot environment
> didn't match the kickstart environment.
I think nics that use the same driver or at least sets on the same card
are detected in the same order. Ours tend to have at least a pair of
Broadcomm and one or more pairs of Intels and the pairs tend to flip
randomly until they are nailed down with a HWADDR= entry in the
ifcfg-eth? file which must rename them after kernel detection.
--
Les Mikesell
lesmikesell at gmail.com