How do I add a 2nd host to a subnet that exists in the 1st host?

[Anticlutch]

I have two hosts.
anti is the first, my primary host. It' s network devices looks like this.
My goal is to add host 2, my new host, to the vmbr1 10.10.10.1/24 subnet.



But I'm not sure how to do this, I don't know how to bridge, or connect the two together.
Both hosts exist in the same datacenter.

 

[cuhanhtuoi]

It looks like there are no ports attached to vmbr1. I recommend adding at least one port before creating a cluster network.
In a Proxmox cluster, you can configure up to 8 cluster networks. When creating the cluster, you can use your vmbr1 network as one of the cluster networks.
You can follow the instructions in the Proxmox documentation for creating a cluster with multiple cluster networks.

# pvecm create CLUSTERNAME --link0 10.10.10.1,priority=30

https://pve.proxmox.com/wiki/Cluster_Manager

 

[Hannes Laimer]

Add as in "add to cluster"(do you already have a cluster? generally 2-node clusters are not stable, you'd probably want a qdevice or third node[1]) or as in "i want guests to communicate over 10.10.10.0/24 with eachother"?

I assume you want to use 10.10.10.0/24 for guests to communicate? If that's the case the simplest way forward is to configure a VXLAN zone[2] under Datacenter > SDN, then add a VNet to it. When you then attach guests to this(select it on the network config of the guest), for guests on the different hosts it'll look like they are directly connected.
If not, please clearify what you mean with in the same datacenter? Are they connected with a cable, and if not how can they reach eachother?

[1] https://pve.proxmox.com/wiki/Cluster_Manager#pvecm_cluster_requirements
[2] https://pve.proxmox.com/pve-docs/chapter-pvesdn.html#pvesdn_zone_plugin_vxlan

 

[Grace]

If both hosts are in the same datacenter, the simplest approach is to connect them through the same Layer 2 network (physical switch/VLAN). Then create a vmbr1 bridge on the second host and attach it to the corresponding physical NIC. Assign the host an unused IP in the 10.10.10.0/24 subnet (e.g. 10.10.10.x/24). If 10.10.10.1 is the only bridge IP on the first host, there's no need to bridge the two Proxmox hosts directly—the physical network handles connectivity.

 

[Anticlutch]

https://pve.proxmox.com/wiki/Cluster_Manager

Thanks, not sure how to add ports to a bridge, I'll read that now.
I 'solved' this by working around PVE and just adding a second static route to my router.


It works, I can ping between the subnets of each PVE host. Unsure of how suitable that will be.
Sad thing is the cost of storage right now.

Add as in "add to cluster"(do you already have a cluster? generally 2-node clusters are not stable, you'd probably want a qdevice or third node[1]) or as in "i want guests to communicate over 10.10.10.0/24 with eachother"?

I assume you want to use 10.10.10.0/24 for guests to communicate?

(by "guest" do you mean containers served by the host?)
If so yes, there are 2 nodes, anti is the first node and pve2 is the 2nd node--each host has a 10.10.__.*/24 subnet, where __ is 10 (anti) or 20 (pve2)
10.10.10.0/24 is a subnet for LXCs served by the anti host who gain their internet through a gateway node also on the anti host.

If not, please clearify what you mean with in the same datacenter? Are they connected with a cable, and if not how can they reach eachother?

As stated above, I simply (manually) duplicated this and used a static route on my router to achieve communication between the hosts anti and pve2. I only need to create a separate node on pve2 that only has access to vmbr1 and see if I have the gateway node working properly. By same datacenter I mean:

The physical machine each have two nics. I could fashion a short cable to directly connect them too. I don't know how I'd achieve communication between the two then, though, as without a static route I'm lost, but I think that's what each of you is hinting at through a VXLan.

through the same Layer 2 network (physical switch/VLAN). Then create a vmbr1 bridge on the second host and attach it to the corresponding physical NIC.

Okay so this is what I use ip addr to see the names of: NIC device names?

 

[Hannes Laimer]

ok, thanks for the information! you already have the two in a cluster, so VXLAN would basically use the connection they already have by encapsulating traffic for the 10.10.10.0/24 network. But since you are able to connect them via a second cable you can also do that and skip SDN, it also avoids guest(yes, thats just the vm or containers you host) traffic from impacting the cluster.

To do that you edit vmbr1 and enter eno1 (on the other host this name might be different, just check, it's the one not assigned to vmbr0) in the "Bridge Ports" field, you don't need to give the hosts an ip on vmbr1 if you don't want your guest to speak with the hosts. So you just connect the two interfaces, then add the interfaces as bridge port to vmbr1 on both. That's it. Guests that are connected to vmbr1 are now connected to each other, and if they have and ip on the same subnet configured they can reach each other. Note there is no switching or routing involved, these are just two (virtual) bridges connected by a cable, so all guests are on the same L2 domain.

Hope this helps!

 

[Anticlutch]

Okay, I think I see where you're going now. I add the device (your ex eno1) as named in the network devices from Type: Network Device, which may be different for each guest/host but I need to make sure to pair them to the physically connected RJ4t ports on the hardware.
host anti:


host PVE2:


So I'd just pick on anti eno1 or enp3s0, figure out which RJ45 port that is and pair it directly with nic0's or nic1's RJ45 port on PVE2... yes?
Then I would use that network bridge and assign its port/slave to the appropriate network device?

e:
I don't know why the name for the second device on anti is so complicated. If I change each of those names to nic0 and nic1, will that propagate through the PVE host (anti), or will I have to go through everywhere and fix it?

ee:
Oh, I can't change the names. Nevermind.