Hi Sven, Frank,

We hit this issue too last year in our CloudStack fork, and this is how we 
resolved it:
https://github.com/MissionCriticalCloudOldRepos/cosmic-plugin-hypervisor-kvm/pull/14

Shouldn’t be too hard to port to current CloudStack, but I have no time to do 
it and test it. We have been running like this ever since (18 months) without 
issues.

Regards,
Remi



On 01/11/2017, 11:06, "Vogel, Sven" <sven.vo...@kupper-computer.com> wrote:

    Hi Frank,
    
    i filed a bug report. I hope anybody can fix that fastly.
    
    https://issues.apache.org/jira/browse/CLOUDSTACK-10127
    
    Greetings
    
    Sven
    
    -----Ursprüngliche Nachricht-----
    Von: Vogel, Sven [mailto:sven.vo...@kupper-computer.com] 
    Gesendet: Mittwoch, 1. November 2017 00:14
    An: dev@cloudstack.apache.org
    Betreff: Re: 4.9 / 4.10 KVM + openvswitch + vpc + static nat / secondary ip 
on eth2?
    
    Hi Frank,
    
    Great analysis, but what can we do now? Is there a workaround or who can 
fix this?
    
    
    Greetings
    
    Sven
    
    
    
    Am 31.10.2017 um 22:41 schrieb Frank Maximus 
<frank.maxi...@nuagenetworks.net<mailto:frank.maxi...@nuagenetworks.net>>:
    
    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
    
    
    On Tue, Oct 31, 2017 at 8:35 PM Frank Maximus < 
frank.maxi...@nuagenetworks.net<mailto:frank.maxi...@nuagenetworks.net>> wrote:
    
    I think that the bug you mentioned might not be related.
    Could you please send the content of the file /etc/cloudstack/ips.json on 
the VR.
    
    That might provide useful information.
    
    Kind regards,
    Frank
    
    On Tue, 31 Oct 2017 15:23 Vogel, Sven, 
<sven.vo...@kupper-computer.com<mailto:sven.vo...@kupper-computer.com>>
    wrote:
    
    Hi Devs,
    
    i have the following problem.
    
    When I look this jira ticket I see no solution.
    https://issues.apache.org/jira/browse/CLOUDSTACK-6801  but I think the 
problem is not solved correctly.
    
       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"
    
    Whats wrong or what can I do to fix this? I hope anybody has an idea.
    
    
    greetings
    Sven Vogel
    Head of Cloud Solutions / Consultants
    Kupper Computer GmbH
    Prager Str. 15
    04103 Leipzig
    
    
    
    

Reply via email to