On Thu, Nov 15, 2018 at 11:50 AM Tony Brian Albers <tba@kb.dk> wrote:
Hi Nathanaël,
Yeah, I think I'll have to just use a physical NIC for the ovirtmgmt. Too bad, it would be nice to be able to separate it from the standard network.
/tony
Actually in several environments I can use globally only 4 network adapters, but they are all 10Gb/s. So in this case is a shame to use a pair of them for ovirtmgmt only... I use 2 of them in bonding (802.3ad) for ovirtmgmt and VMs, and 2 of them for iSCSI, migration network and NFS access The culprit is to pre-define the bond on the host and set ovrtmgmt as a vlan based network with the vlan id that is necessary in your case. Eg on one environment of these my vlan for ovirtmgmt is 68 In this case on the host there will be bond0 before setup and then after setup they will be created bond0.68, bond0.65 interfaces and ovirtmgmt bridge and vlan65 bridge 8: bond0: <BROADCAST,MULTICAST,MASTER,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000 link/ether 18:03:73:0d:e9:41 brd ff:ff:ff:ff:ff:ff 24: bond0.65@bond0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue master vlan65 state UP group default qlen 1000 link/ether 18:03:73:0d:e9:41 brd ff:ff:ff:ff:ff:ff 25: vlan65: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000 link/ether 18:03:73:0d:e9:41 brd ff:ff:ff:ff:ff:ff 26: bond0.68@bond0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue master ovirtmgmt state UP group default qlen 1000 link/ether 18:03:73:0d:e9:41 brd ff:ff:ff:ff:ff:ff 27: ovirtmgmt: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000 link/ether 18:03:73:0d:e9:41 brd ff:ff:ff:ff:ff:ff inet 10.4.192.32/24 brd 10.4.192.255 scope global ovirtmgmt valid_lft forever preferred_lft forever [root@ov200 ~]# brctl show bridge name bridge id STP enabled interfaces ;vdsmdummy; 8000.000000000000 no ovirtmgmt 8000.1803730de941 no bond0.68 vlan162 8000.a0369f2e1388 no p1p1.162 vlan65 8000.1803730de941 no bond0.65 vnet0 vnet2 vlan98 8000.a0369f2e138a no p1p2.98 vnet1 vnet3 [root@ov200 ~] with the ovirtmgmt bridge on bond0.68 device In another environment I have 4 interfaces and ovirtmgmt not set up as vlan based, but I can configure the switch ports as trunk with vlan 187 enabled and as default vlan set the 68 one, so that if packets are not tagged they go through that vlan. In this case on one host I have to preconfigure the bond0 and then at the end of setup I will have something like this: 45: bond0: <BROADCAST,MULTICAST,MASTER,UP,LOWER_UP> mtu 1500 qdisc noqueue master ovirtmgmt state UP group default qlen 1000 link/ether f4:e9:d4:d6:15:a0 brd ff:ff:ff:ff:ff:ff 46: ovirtmgmt: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000 link/ether f4:e9:d4:d6:15:a0 brd ff:ff:ff:ff:ff:ff inet 10.4.192.47/24 brd 10.4.192.255 scope global ovirtmgmt valid_lft forever preferred_lft forever inet6 fe80::f6e9:d4ff:fed6:15a0/64 scope link valid_lft forever preferred_lft forever 47: bond0.187@bond0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000 link/ether f4:e9:d4:d6:15:a0 brd ff:ff:ff:ff:ff:ff inet 10.4.187.102/24 brd 10.4.187.255 scope global bond0.187 valid_lft forever preferred_lft forever 48: bond0.65@bond0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue master vlan65 state UP group default qlen 1000 link/ether f4:e9:d4:d6:15:a0 brd ff:ff:ff:ff:ff:ff 49: vlan65: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000 link/ether f4:e9:d4:d6:15:a0 brd ff:ff:ff:ff:ff:ff and # brctl show bridge name bridge id STP enabled interfaces ;vdsmdummy; 8000.000000000000 no ovirtmgmt 8000.f4e9d4d615a0 no bond0 vlan65 8000.f4e9d4d615a0 no bond0.65 vnet0 vnet1 So the ovirtmgmt bridge is on bond0 and the VM dedicated bridge (vlan65) is on bond0.65 And on bond0.187 I have the ip dedicated to live migration. The important thing is to presetup the bond and make the right choice at the beginning about ovirtmgmt, if you want ovirtmgmt vlan based or not. And also configure accordingly the switch ports, of course. HIH, Gianluca