Re: upstream systemd discussion about MACAddressPolicy for bonds and bridges
On 10/05/2022 14:57, Dusty Mabe wrote:
On 5/10/22 02:05, Tom Hughes via devel wrote:
> On 10/05/2022 03:12, Dusty Mabe wrote:
>
>> Just wanted to point interested people in the direction of an upstream
>> discussion about how (by default) the MAC address should get set for
>> bond and bridge devices.
>>
>> https://lists.freedesktop.org/archives/systemd-devel/2022-May/047893.html
>>
>> A few of us were originally going to propose a change to Fedora to change
>> the current behavior back, but it was suggested we take the discussion
>> upstream to hash out the merits there.
>
> All I can say is, no, just leave it alone.
>
> Having fixed all my machines two years ago when it stopped generating
> persistent addresses I have just spent this weekend doing it again now
> that F36 has gone back to them.
I'm not aware of any change in behavior in F36, but maybe I missed something.
F36 has gone back to using a persistent MAC calculated from
the machine ID and bridge name.
I'm not sure what F35 was doing but believe me when I say that
all my bridges changed MAC on upgrade and they've now gone back
to the address they had until late 2019 or early 2020.
> I don't care what address my bridges have, so long as it is
fixed so
> that DHCP can allocate addresses against it, but I do prefer not to
> have to fix all my DHCP and DNS every time the policy flip flops and
> upgrades break my networks!
With the current policy you'll get a new MAC every time you re-install
a machine. Is that what you want?
It doesn't bother me too much as I expect to be reconfiguring
things then.
The upstream proposal is to make it based on the actual MAC of the
NIC(s)
again.
The problem is which NIC exactly? As I understand it's whatever
interface gets added first so may be racy.
From that point of view a persistent address seems better.
Tom
--
Tom Hughes (tom(a)compton.nu)
http://compton.nu/