Possibility defining uniqueness for MAC addresses on subnets, not hosts

I'm posing this question/feature request as a way to determine whether this
would even be acceptable option for Foreman, or if the thought is enough to
make Foreman devs angry. :slight_smile:

For some context: within multiple datacenters, a user may manage multiple
subnets with the possibility for identical mac addresses on interfaces
within those layer 2 networks. As far as I'm aware, this doesn't violate
any RFC per se - globally unique identifiers are still being maintained
within a network node segment, just not across layer 3.

The use case for this is varied - for example, many hosts can have their
virtual MAC addresses defined by scripts in a way that makes for consistent
addresses across datacenters, which allows hardware to be quickly swapped
out for new hardware that will acquire predefined addresses.

Thus while I do think that uniqueness for MAC addresses should be
maintained in Foreman, it would make life easier for some datacenter
designs to place the restriction on a NIC within a subnet (where you
obviously wouldn't want duplicate MACs) and not hosts globally.

Let me know if this sounds reasonable or if I'm completely off-base here.

I can see your point, which seems reasonable to me, though there are a
couple of things that might break:

  1. Kickstarts take a parameter called kssendmac, which includes the MAC
    address in the header of the kickstart (unattended provision template)
    request. Foreman can use this to identify the host (else a token or IP
    match). Since tokens are enabled by default now, this is less of an issue.

  2. The bootdisk plugin works similarly when using the generic ISO image.

You'd probably also need some uniqueness per TFTP smart proxy, which
makes it more complex - consider the situation where a user creates two
subnets, but with the same TFTP proxy assigned.

Maybe there are some more gotchas I haven't thought of…

··· On 24/06/14 18:46, tylerl@qualtrics.com wrote: > I'm posing this question/feature request as a way to determine whether > this would even be acceptable option for Foreman, or if the thought is > enough to make Foreman devs angry. :) > > For some context: within multiple datacenters, a user may manage > multiple subnets with the possibility for identical mac addresses on > interfaces within those layer 2 networks. As far as I'm aware, this > doesn't violate any RFC per se - globally unique identifiers are still > being maintained within a network node segment, just not across layer 3. > > The use case for this is varied - for example, many hosts can have their > virtual MAC addresses defined by scripts in a way that makes for > consistent addresses across datacenters, which allows hardware to be > quickly swapped out for new hardware that will acquire predefined addresses. > > Thus while I do think that uniqueness for MAC addresses should be > maintained in Foreman, it would make life easier for some datacenter > designs to place the restriction on a NIC within a subnet (where you > obviously wouldn't want duplicate MACs) and not hosts globally. > > Let me know if this sounds reasonable or if I'm completely off-base here.


Dominic Cleal
Red Hat Engineering