Remi, We haven't experienced exactly what you're describing, but we did have a similar issue that caused multiple interfaces to be built back when we upgraded to 4.3 from 4.1.1 on KVM. Have a look at https://issues.apache.org/jira/browse/CLOUDSTACK-6464. The notes also reference another couple of issues that might have been related.
This particular issue was apparently fixed in 4.4, but nevertheless, it might be worth a look to see if anything jumps out for you. - Si ________________________________________ From: Remi Bergsma <rberg...@schubergphilis.com> Sent: Tuesday, April 14, 2015 4:19 PM To: dev@cloudstack.apache.org Subject: KVM plugs public nic twice in non-VPC router Hi, We’re testing KVM and noticed that a virtual router (non-VPC) gets two pubic ip nics plugged. This is in 4.4.2 when using Open vSwitch on CentOS 7 (advanced zone). The reason for this, is that the first nic gets plugged by provisioning the nic. Then, a second nic is plugged that gets the same ip address but a slightly different nic. This is done by IpAssocCommand. When diving into this, this seems to happen because the StartCommand contains a IpAssocCommand part with again the public ip. This is probably because this is how we plug the public nic on VPCs. I think we should remove it for non-VPC routers. Here you see the second public nic getting plugged: 2015-04-14 21:19:20,099 DEBUG [cloud.agent.Agent] (agentRequest-Handler-5:null) Processing command: com.cloud.agent.api.routing.IpAssocCommand 2015-04-14 21:19:20,103 DEBUG [kvm.resource.OvsVifDriver] (agentRequest-Handler-5:null) plugging nic=[Nic:Public-null-vlan://50<public-null-vlan://50>] Notice the ip-address part of the object is empty. This leads to this command being ran on the router: Apr 14 19:19:20 r-31443-VM cloud: VR config: executing: /opt/cloud/bin/ipassoc.sh -A -s -f -l 1.2.3.4/24 -c eth3 -g 1.2.3.1 -n >From a functional perspective this does work, but is not what it should be if >you ask me. It looks like this issue: https://issues.apache.org/jira/browse/CLOUDSTACK-4690 But the story about many many nics I cannot reproduce. I do nowever found that associating extra public ip addresses does not work. The management server reports everything fine, but I see nothing happening on the router nor the agent on the hypervisor: 2015-04-14 23:09:13,437 DEBUG [cloud.network.IpAddressManagerImpl] (API-Job-Executor-82:ctx-f4953f35 job-534671 ctx-2a536669) Associating ip Ip[1.2.3.5-2] to network Ntwk[a4960690-4a29-445e-b745-881c13bca270|Guest|14] 2015-04-14 23:09:13,449 DEBUG [cloud.network.IpAddressManagerImpl] (API-Job-Executor-82:ctx-f4953f35 job-534671 ctx-2a536669) Successfully associated ip address 1.2.3.5 to network Ntwk[a4960690-4a29-445e-b745-881c13bca270|Guest|14] Question: Does anyone else with KVM running see this behaviour? Or am I missing something? Thanks, Remi
