control: retitle -1 'grml-debootstrap: support predictable network device names in VM images'
* Holger Levsen [Mon Jul 06, 2020 at 10:27:32AM +0000]: > On Wed, Feb 21, 2018 at 02:58:18PM +0100, Holger Levsen wrote: > > i'm trying to setup stretch images, and while the same commandline works > > for jessie, it fails with stretch like this: [...] > long story short, this ^ is the cause for what I have seen, using stretch > and buster the network interface is 'ens3' and not 'eth0', thus the network > didnt come up as it should... > Now that I know this, I'm using 'ens3' for these systems but I'd rather like > grml-debootstrap to (optionally) append 'net.ifnames=0' to the kernel > commandline, thus forcing the interface name to be 'eth0'. I've changed the > bug meta-data accordingly. Thanks for providing an update to the bug report and clarifying this! JFTR: we already have support for predictable network interfaces (see e.g. --defaultinterfaces), though it indeed is broken as-is for the VM image use case, especially as the network device name(s) depend(s) on the runtime environment of the VM image (and e.g. ifupdown sadly isn't as dynamic as someone might wish). So it might indeed make sense to enable net.ifnames=0 by default in the VM image use case. If anyone should have a better idea, please let me/us know! regards -mika-
signature.asc
Description: Digital signature
