[Server-devel] Anaconda install - conflicts over ifcfg-eth0

[Jerry Vonau]

Jerry Vonau wrote:
> Martin Langhoff wrote:
>> Executive version: I am preparing a package that provides an
>> ifcfg-eth0 file - and finding that Anaconda overwrites it...
>>
>> If the package says %config{noreplace}
>> /etc/sysconfig/network-scripts/ifcfg-eth0 then anaconda installs the
>> file as an .rpmnew file. On the other hand, if I remove the
>> '{noreplace}' option, anaconda writes an .rpmorig file with its
>> settings _and also_ overwrites the ifcfg-eth0 file. The result is that
>> cmp ifcfg-eth0{,.rpmorig}  is true.
>>
>> This looks to me like an anaconda bug :-/ Are there reasonable ways to
>> make this work? Can I tell anaconda that the network config used
>> during install is _not_ to be saved permanently?
>>
>> As a workaround, I can overwrite the file on %post but that gets nastier...
>>
>> Note! This package is not meant for Fedora proper, naturally. It is a
>> configuration package for the OLPC School Server, with (mostly) good
>> reasons to do the wild things it does :-)
> 
> Anaconda by default doesn't create an ifcfg-eth0 file, network-manager 
> doesn't need one to work. 
Sigh, the livecd boots without an ifcfg-eth0 file, sorry long day..

> Are you using a kickstart file? Sounds like 
> the net config is coming from that.
>

The test dvd installed the ifcfg-eth0 with the options(dhcp here)
from the ks.file that I used. The ifcfg-eth0.rpmnew has the goods.

Maybe just have network_config fix the .rpmnew issue if needed, if the 
XS_SERVER_NUM is not set, that would be the clear clue to overwrite the 
stock ifcfg file, IMHO. Still would need to run domain_config anyway, so 
it's not like everything can run out of the box on first boot...

Jerry