Description of problem: Version-Release number of selected component (if applicable): How reproducible: allways Steps to Reproduce: 1. have a machine where PXE boot is done from eth1, eth0 is disconnected 2. PXE boot the machine for NFS kickstart 3. Actual results: After the kernel hass been downloaded from eth1 (and initrd), anaconda tries to get a DHCP address for the non-connected eth0. If you plug in a cable to eth0 just to get it an address, anaconda continues to kickstart from eth1 (as expected using ksdevice=bootif and IPADDRESS 2 in the pxe config file Expected results: eth0 should be left alone, eth1 should have been used Additional info: this is a desktopmachine with an Intel Pro1000 interface on board with PXE capabilities. An additional (8139too) ethernet interface is installed in one of the PCI interfaces. The BIOS will use the intel card to boot, but then the kernel sees the intel card as eth1, and the realtek card as eth0
Created attachment 290027 [details] lspci -v output of system to be kickstarted
Created attachment 290028 [details] output of ifconfig -a on system
Created attachment 290029 [details] relevant part of pxe config file
What does your kickstart file look like? If you have network commands for eth0 and eth1, then anaconda will try to bring up all of those interfaces as specified. When doing a kickstart install, we bring up all interfaces listed in the kickstart file because the user may have something in %pre or %post that depends on the other interfaces being up.
Created attachment 304601 [details] a relevant kickstartfile is here.
What you are describing is expected. You have two network lines in your kickstart file, so anaconda tries to bring up both interfaces during installation. Your description also states that ksdevice=BOOTIF is working, it's just that eth0 is brought up anyway...which is because you have two network lines in your kickstart configuration file.