Node network configuration

Each node’s networking is configured under node → System → Network. This is where the bridge that guests attach to lives, along with any bonds and VLANs beneath it.

What a new node looks like

Name	Type	Ports	Address
enp5s0	Network Device		(none)
vmbr0	Linux Bridge	enp5s0	the management address

The physical adapter carries no address of its own; it is a port on a bridge, and the bridge holds the address. Machines attach to the bridge, which is why they appear on the same network as the node.

Changes are staged, not immediate

This panel does not apply changes as you make them. Edits are held as a pending configuration and only take effect when you click Apply Configuration. Revert discards everything pending.

That is deliberate: you can build a whole new bond and bridge, review it, and commit once.

Two ways this catches people out

An edit that is never applied does nothing at all — the interface will look configured and behave as though it is not.

And an edit that is applied takes effect immediately, including on the interface you are connected through. A wrong address or a bridge with the wrong port attached ends your session, and recovering means getting to the server’s own console.

Before applying a change to the management interface, make sure you have another way in — console access, or someone physically present.

The interface types

Create offers six:

Type	What it is for
Linux Bridge	A virtual switch. What machines attach to.
Linux Bond	Several physical adapters combined for redundancy or throughput.
Linux VLAN	One VLAN split out of an underlying interface.
OVS Bridge, OVS Bond, OVS IntPort	Open vSwitch equivalents

Use the Linux types unless you specifically need Open vSwitch.

Linux Bridge

Field	Notes
Name	Conventionally vmbr followed by a number
IPv4/CIDR, Gateway (IPv4)	Leave empty for a bridge machines use but the node does not
Bridge ports	The interface beneath it — a physical adapter, a bond, or a VLAN
VLAN aware	Lets machines set their own VLAN tags on this bridge
Autostart	Bring it up at boot. Normally on.
MTU	Only if your network uses non-standard frame sizes

Decide “VLAN aware” up front

A VLAN-aware bridge lets each machine pick its VLAN in its own network settings, instead of needing a separate bridge per VLAN. Changing it later disrupts everything on the bridge, so choose at creation.

Linux Bond

Field	Notes
Slaves	The physical adapters being combined
Mode	How traffic is distributed and how failover works
Hash policy	How traffic is spread across members, for the modes that spread it
bond-primary	The preferred adapter in active-backup mode

The bond mode must match the switch

Modes that aggregate links need the switch configured to match. A mismatch usually does not fail cleanly — it produces intermittent loss that is hard to trace. If you are not sure what the switch is doing, active-backup works with any switch and still survives a cable or port failure.

Linux VLAN

Field	Notes
Vlan raw device	The interface to take the VLAN from — often a bond
VLAN Tag	The VLAN ID

A typical production layout

For a server with two network ports and a tagged switch link:

1. Create → Linux Bond — both physical adapters as Slaves, a mode your switch supports. No address.
2. Create → Linux VLAN — raw device the bond, tag the management VLAN.
3. Create → Linux Bridge — bridge port the VLAN interface, management address and gateway on the bridge.
4. Repeat 2–3 for each additional VLAN machines need.
5. Review, then Apply Configuration.

This gives redundant physical links, separated networks, and one bridge per network for machines to attach to.

Give cluster traffic its own network if you can

Cluster membership depends on small messages arriving predictably. On a busy shared network they can be delayed enough for a healthy node to be treated as failed. A separate network — even a modest one — avoids that. See Clustering.

Related settings

Setting	Where
Search domain and name servers	node → System → DNS
Static name-to-address entries	node → System → Hosts
Virtual networks for guests	Datacenter → SDN — see Software-defined networking

Cluster members must resolve each other

Where there is no reliable DNS on the management network, the entries under Hosts are what let nodes find one another. A node that installs cleanly but will not join a cluster very often has a name that does not resolve.

If something goes wrong

What you see	What to do
Changes had no effect	They are still pending. Click Apply Configuration.
You lost access after applying	Get to the server’s console to correct it. There is no way back through the interface you just broke.
A machine has no network	Check which bridge its adapter is attached to, and that the bridge has a port.
A VLAN-tagged machine cannot reach anything	The bridge is not VLAN aware, or the switch port does not carry that VLAN.
A bond gives redundancy but no extra throughput	Expected in active-backup. Aggregating modes need matching switch configuration.

Still stuck? Contact VM2Cloud support.