On Tue, Oct 27, 2015 at 9:21 AM, Michal Kubecek
On Tuesday 27 of October 2015 06:34:30 Andrei Borzenkov wrote:
The problem with persistent was renaming inside the same namespace that was not reliable. People were actually left with eth0_temp at times.
This has been fixed.
eth0 wants to be renamed to eth1. eth1 gets renamed to eth1_temp; at the same moment new interface is detected that gets eth1 from kernel. Now you can neither rename eth0 to eth1 nor rename eth1_temp back. Current net_id gets it right by making each event independent - you only need to care about current interface which generated event.
Another consideration is that IMHO persistent renaming based on MAC is wrong in most cases and should have never been default. This is fixable again.
There are special scenarios when it's not suitable (once I've seen a Hyper-V setup which assigned VM's NIC different MAC address each time) but I certainly wouldn't call these "most cases".
This makes hardware replacement much more difficult and demanding than it could be, especially when the only access to system you have remotely is via interface that is replaced. Using topology (device tree if you like) for persistent names indirectly maps interfaces to external connectors which is what most people are really interested in. When you send someone to plug cable into interface, do you tell the person MAC address or physical port location in device? Current udev actually give you full flexibility on selecting which naming scheme for which device is used. -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org