Details
Critical Description
We have the following Problem.
1. KVM
2. Bridges
bond with two interfaces and trunk (0,129,180,100-1500) to cloudbr0
Cloudbr0 (0 - guest network)
Fakebridge pub129 (public network)
Fakebridge sto180 (secondary storage network)
Fakebridge mgmt0 (management)
If I have a vpc all things work until I add a secondary ip and add a static nat.
The following will happen, first address will be on the the correct interface 146.0.122.134/26 but static nat will be on the false network.
Its on the eth2…
{{ root@r-29-VM:~# ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 16436 qdisc noqueue state UNKNOWN
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
inet 127.0.0.1/8 scope host lo
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
link/ether 0e:00:a9:fe:03:81 brd ff:ff:ff:ff:ff:ff
inet 169.254.3.129/16 brd 169.254.255.255 scope global eth0
3: eth1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
link/ether 1e:00:2c:00:00:68 brd ff:ff:ff:ff:ff:ff
inet 146.0.122.134/26 brd 146.0.122.191 scope global eth1
4: eth2: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000
link/ether 02:00:57:07:00:0c brd ff:ff:ff:ff:ff:ff
inet 192.168.1.254/24 brd 192.168.1.255 scope global eth2
inet 146.0.122.135/26 brd 146.0.122.191 scope global eth2}}
Normally I think the secondary ip should be on signed to eth1 not eth2!
It sets my ip on the guest network vlan range on my cloudbr0 but it should be pub129. vnet6 has 1353 guest tag and not the public tag.
[root@kvm01 ~]# ovs-vsctl list-br
cloud0
cloudbr0
mgmt0
pub129
sto180
[root@kvm01 ~]# virsh domiflist r-29-VM
Interface Type Source Model MAC
-------------------------------------------------------
vnet4 bridge cloud0 virtio 0e:00:a9:fe:03:81
vnet5 bridge pub129 virtio 1e:00:2c:00:00:68
vnet6 bridge cloudbr0 virtio 02:00:57:07:00:0c
Bridge "cloud0"
Port "vnet4"
Interface "vnet4"
Port "vnet5"
tag: 129
Interface "vnet5"
Port "vnet6"
tag: 1353
Interface "vnet6"
root@r-29-VM:~# cat /etc/cloudstack/ips.json {
"eth0": [
],
"eth1": [
],
"eth2": [
,
,
{ "add": true, "broadcast": "192.168.1.255", "cidr": "192.168.1.254/24", "device": "eth2", "gateway": "192.168.1.254", "netmask": "255.255.255.0", "network": "192.168.1.0/24", "nic_dev_id": "2", "nw_type": "guest", "one_to_one_nat": false, "public_ip": "192.168.1.254", "size": "24", "source_nat": false } ],
"id": "ips"
}
Frank Maximus from Nuage analysed the problem.
That seems to be a bug in the lookup of the device number, in case of openvswitch.
The config clearly sets device to eth2, while it should be eth1.More specifically:
in LibvirtComputingResource.prepareNetworkElementCommand()
The broadcastUriToNicNum map is filled depending on the VR nics.
In openvswitch the guest bridge is used as is, so it overwrites the mapping of public.
This was not an issue until 4.6 as then VR was using the macaddress to do lookup, while now it is using the device number.Kind Regards,
Frank
I hope anyone can fix that fastly.
Attachments
Issue Links
- links to