Re: [metallb-users] Digest for metallb-users@googlegroups.com - 1 update in 1 topic

[Robin Smith]

I believe that would be mostly because these questions are already answered by L2 protocols.

On Wed, 24 Dec 2025, 01:15 , <metall...@googlegroups.com> wrote:

metall...@googlegroups.com

Google Groups

Topic digest
View all topics

• Is there a reason metallb does not announce on L2 by adding the IP address to the kernel? - 1 Update

Is there a reason metallb does not announce on L2 by adding the IP address to the kernel?

Philip Couling <cou...@gmail.com>: Dec 23 03:29AM -0800 I'm curious about the design of metallb in implementing L2 protocols for itself. What I'm wondering is why it didn't advertise the address by adding and removing the addresses on the host.   I'm asking because I'm interested in implementing an alternative announcer which does just that and I'm wondering if there's a known reason why this is a very bad idea.   From what I can understand this approach would:   - help visibility: - showing sys-admins directly which IPs a node thinks it has responsibility for (shown through the "ip address" command) - responding to ICMP requests allowing services to be pinged - help avoid collisions where metallb doesn't have sole ownership of the CIDR range: - DHCP servers can ping addresses prior to allocating *(RFC 2131 section 3.1)* - Leveraging the kernel's implementation of IPv6 Duplicate Address Detection (DAD) So I'm really interested to understand if there was a reason for not implementing it this way?

Back to top

You received this digest because you're subscribed to updates for this group. You can change your settings on the group membership page. To unsubscribe from this group and stop receiving emails from it send an email to metallb-user...@googlegroups.com.