Multiple PVE (Single-Node) Clusters with VMs and LXCs on Same Subnet: Alter NIC MAC Prefix Per Cluster?

[SInisterPisces]

(I'll preface this by saying I'm in a home server/homelab environment, so I realize this is probably not the correct enterprise production way to run things.)

Anybody out there running multiple Proxmox clusters (single-node or otherwise) with VM and LXC containers whose vNICs share subnets?

I'm trying to figure out if I need to give each cluster its own vNIC MAC prefix under Datacenter -> MAC address prefix to avoid conflicts on the subnet when I'm generating new VMs and LXC containers.

In my case, I've got at least two single-node clusters that I plan to manage under a single pane of glass with Proxmox Datacenter Manager.

I use DHCP with static IP reservations, so it would be a problem if there were MAC address conflicts on VMs and LXCs. (It's late and I'm tired; please don't suggest switching to fully static IP management.

)

I don't think I'll somehow run out of MAC addresses if I don't change them (there's a lot of octet combinations), but there's still a non-zero chance of collisions across clusters and I don't want to confuse my DHCP server.

If I do need to change them, what's the easiest way? Can I just increment the third octet in the MAC prefix by one?

Thanks.

 

[bl1mp]

Well, using different prefixes would be the way to ensure the absence of MAC address conflicts.

Doubled macaddresses can cause network trouble independent of DHCP.

The good news is, if you haven't thought about this, the statistic probability of collisions is very, very low.

BR, Lucas

PS: originally the first bits of a Macaddress, belongs to a specific vendor. I am not sure,
if every manufacturer still follows that.
So there might be ranges more suitable, than others for your usecase

 

[SInisterPisces]

Well, using different prefixes would be the way to ensure the absence of MAC address conflicts.

Doubled macaddresses can cause network trouble independent of DHCP.

The good news is, if you haven't thought about this, the statistic probability of collisions is very, very low.

BR, Lucas

PS: originally the first bits of a Macaddress, belongs to a specific vendor. I am not sure,
if every manufacturer still follows that.
So there might be ranges more suitable, than others for your usecase

Thanks!

I'll have to google a bit and see if I can find some safe prefix ranges, I guess.

 

[shanreich]

You probably want a locally administered unicast MAC address - so you can use any prefix with the second nibble being 2, 6, A, E.

 

[SInisterPisces]

You probably want a locally administered unicast MAC address - so you can use any prefix with the second nibble being 2, 6, A, E.

Thanks! I was googling around a bit and found similar information, but it was all on random forums and message boards from 10-25 years ago, and (of course) people were arguing over what counted as valid or not. I appreciate the confirmation.

I found this Wiki page. I'm going to grab a cup of tea and stare at it until I feel marginally less stupid about this.
https://en.wikipedia.org/wiki/MAC_address#IEEE_802c_local_MAC_address_usage

I also found a couple of generators, but I'm not sure how reliable they are yet.
https://www.hellion.org.uk/cgi-bin/randmac.pl
https://macaddresslookup.io/mac-address-generator

I hate to admit that I still struggle quite a bit with hexadecimal. Every time it comes up, it's confusing, but it doesn't come up often enough for me to get good at it. -_-

EDIT: Just to make sure I actually understand this, I settled on F2 for the first byte. That should be safe, I think. After getting so confused when I first read about it, that almost feels too easy.

 

[shanreich]

EDIT: Just to make sure I actually understand this, I settled on F2 for the first byte. That should be safe, I think. After getting so confused when I first read about it, that almost feels too easy.

Yes, that should be fine!