WAN & LAN / eth0/eth1

[MrMourik]

My testing lab cluster:
3 nodes, 1 SAN, connected with FiberChannel & LVM

So far, so good.

I tried today migrating KVM's over the node's, and this working GREAT, BUT...

At this moment I just configured ONE ethernet connection on each node (all eth0/vmbr0)
and so, all the migrating traffic is going over that interface

When i put everything in production, i want to seperate traffic:

1. Internet / WAN
2. Internal / LAN (migrating)
3. FC for the image's (all done)

Is this possible? Can I just setup eth1/vmrb1 as LAN traffic?

I thougt I read somewhere that migrating is only possible on vmbr0 !

So maybe it's better to setup :

eth0/vmbr0 = LAN (192.168.100.0/24)
eth1/vmbr1 = WAN/Internet (213.x.x.x/27)

 

[udo]

My testing lab cluster:
3 nodes, 1 SAN, connected with FiberChannel & LVM

So far, so good.

I tried today migrating KVM's over the node's, and this working GREAT, BUT...

At this moment I just configured ONE ethernet connection on each node (all eth0/vmbr0)
and so, all the migrating traffic is going over that interface

When i put everything in production, i want to seperate traffic:

1. Internet / WAN
2. Internal / LAN (migrating)
3. FC for the image's (all done)

Is this possible? Can I just setup eth1/vmrb1 as LAN traffic?

I thougt I read somewhere that migrating is only possible on vmbr0 !

So maybe it's better to setup :

eth0/vmbr0 = LAN (192.168.100.0/24)
eth1/vmbr1 = WAN/Internet (213.x.x.x/27)

Hi,
to 3:
FC has nothing to do with IP - so you have own fc-adapter. If you have iSCSI you can use an own NIC (if the network is used for iSCSI only you don't need an bridge).

to 1 and 2:
right, it's makes sense to split the networks, that vmbr0 is the local one.
You can name the bridges like the network (except vmbr0) to get an better overview - like vmbr213 for 213.x.x.x.

I do it in same matter - vmbr0 only for clustercommunication and SAN (iSCSI/DRBD) on one 10GB-Nic and on the other 10GB-Nic all other networks with vlan tagging (vmbr20 for x.x.20.0/24).
BTW: Only the bridges how need access from that network to the pve-gui (and where the normal routing will fail) need an IP - and of course vmbr0.


Udo