Plugins » Discovery

More on the topic:

https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/html/Networking_Guide/ch-Consistent_Network_Device_Naming.html

https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/html/Networking_Guide/sec-Disabling_Consistent_Network_Device_Naming.html

http://cgit.freedesktop.org/systemd/systemd/tree/src/udev/udev-builtin-net_id.c#n20

It looks like old names are still available in the kernel log:

$ dmesg | grep rename
[    1.935439] e1000e 0000:00:19.0 em1: renamed from eth0
[    4.640024] iwlwifi 0000:03:00.0 wlp3s0: renamed from wlan0

Therefore it could be possible to write a custom fact extension to send these and then to modify discovery plugin to be able to use the old names. This will likely need a change in provisioning forms tho.

I got further in my installs and I see that CentOS 6 is simply renaming ethx to emx and next time puppet runs, most of the foreman interface layout is updated, but it sure is goofy.

The problem i had with it asking for the install interface turned out to be that anaconda wasn't aware of vlan tagging, so i added --vlanid to the bootproto of the kickstart file and now machines boot with their tagged vlan/primary interface which i think may have been the real issue here..

Either way though, the process is kludgy with the names changing - especially if you try and write puppet code to handle people making changes and having it converge through the next puppet run since I see some servers boot with em1-em3 and then eth0 is left out there as a udev device.

Doing a lot more experiments on bare metal to see if i can replicate the failures with log files

Lukas, what's the status here?