Hello.
Thank you for the pointer!
Manually defining the bridge in netplan seems to have brought it up and
SystemVMs are now recognized by Cloudstack.
The SystemVMs can also now be SSH-d into with link-local address.
When looking at cloudbr0 with ovs-vsctl, all ports are untagged. The other,
cloudbr1 is tagging the networks properly.
We have bond0 configured with untagged going to management and storage should
go through a tagged network.
However, when trying to define the VLAN/VNI in Cloudstack interface, it gets
removed when IP range is added.
Maybe we need to do a cloudbrXX with the storage vlan ID as XX?
Topology:
- LOM NIC: eno5np0/eno6np1 - Management/Storage. LACP, Trunk port with
Management and Storage VLAN.
- Separate 40G NIC: ens2/ens2d1 - Guest/Public traffic. LACP, Trunk with
Guest/Public VLANs.
Inside SSVM, due to missing tag, it cannot access secondary storage.
eth0 and eth1 function as expected.
eth2/eth3 claim to have no route to host.
The netplan config with some internal info redacted/changed is as follows:
```
network:
version: 2
ethernets:
eno5np0: {}
eno6np1: {}
ens2: {}
ens2d1: {}
bonds:
bond0:
interfaces: [eno5np0, eno6np1]
openvswitch:
lacp: active
parameters:
mode: balance-slb
lacp-rate: slow
mii-monitor-interval: 100
bond1:
interfaces: [ens2, ens2d1]
openvswitch:
lacp: active
parameters:
mode: balance-slb
lacp-rate: slow
mii-monitor-interval: 100
bridges:
cloudbr0:
interfaces: [bond0]
addresses: [192.168.10.11/24]
routes:
- to: default
via: 192.168.10.1
nameservers:
addresses: [*.*.*.*, *.*.*.*]
openvswitch: {}
cloudbr1:
addresses: []
interfaces: [bond1]
openvswitch: {}
cloud0:
dhcp4: no
addresses: [169.254.0.1/16]
openvswitch: {}
```
Lugupidamisega / Best regards,
Joann Mõndresku
Süsteemiadministraator | Systems Administrator | [email protected]
WaveCom AS | ISO 9001, 27001 & 27017 Certified DC and Cloud services
Endla 16, Tallinn 10142 | www.wavecom.ee | www.facebook.com/wavecom.ee
----- Original Message -----
From: "Wei ZHOU" <[email protected]>
To: "users" <[email protected]>
Sent: Thursday, November 28, 2024 3:08:35 PM
Subject: Re: OVS + CloudStack on Ubuntu22
I have faced the issue before that cloud0 is DOWN after the host is
rebooted. Before the host is rebooted, cloud0 is Up (brought up by
cloudstack agent)
It seems to be because cloud0 does not have any interface associated with
it.
my workaround is letting netplan creating the bridge
cloud0:
dhcp4: no
addresses: ["169.254.0.1/16"]
openvswitch: {}
I did not face the issue with your storage network, it's better to share
your netplan configuration.
-Wei
On Thu, Nov 28, 2024 at 11:47 AM Joann Mõndresku <[email protected]>
wrote:
> Hello.
>
> We are in process of setting up a POC environment with KVM and OVS on
> CloudStack.
> However, after adding the host to CloudStack, the automatically created
> cloud0 interface is down and none of the systemvm link-locals work.
> Tagged interfaces do not seem to want to work either.
>
> Our storage network is also on a separate Vlan, when trying to define
> VLAN/VNI field, it gets erased on save.
>
> Anyone with successful deployment of KVM/OVS setup have any pointers on
> where I could have gone wrong here?
> Is it even a recommended/widely used setup on KVM?
>
>
> Lugupidamisega / Best regards,
>
> Joann Mõndresku
>
>
>
> Süsteemiadministraator | Systems Administrator | [email protected]
>
> WaveCom AS | ISO 9001, 27001 & 27017 Certified DC and Cloud services
> Endla 16, Tallinn 10142 | www.wavecom.ee | www.facebook.com/wavecom.ee
>
