Hi there!
We want to change few thing is our production Openstack setup around
networking. (it's Mitaka now, but soon we planning the upgrade)
Just a few idea, how it looks now:
- We use OVS
- Have 2 network nodes (l3 agent, dhcp agent, etc)
- Have multiple compute nodes
- Have independent controllers (with neutron-api)
You can choose from two network type - please ignore the naming now - :
- "Flat"
- "Smart"
The "Smart" dataflow looks like this: backbone -> network nodes (where
the kernel get down the VLAN tag, and add the untagged traffic to
br-smart) OVS/qrouter -> vxlan tunnel -> compute node OVS -> VM
The "Flat" dataflow looks like this: backbone -> compute node (same as
network node, we have bondX.100, bondX.101 etc interfaces) -> OVS bridge
(br-flat) -> OVS - > VM (the network nodes also have interface in our
"flat" networks, since they doing dhcp/metadata, but the gateway is two
HW router with vrrp, not some dynamic qrouter.
Of course, there are firewalls before the VM, but it's irrelevant.
So, our goal is to unconfiugre this bridge mass, and do the vlan
tagging/untagging in OVS.
AFAIK it can work, if I make a trunk port in OVS, and then choose VLAN
type with the right segmentation ID. (I'am possibly okay with the
configuration itself, the next part looks hard)
So, as you can imagine, we have multiple VM-s in "smart" and "flat"
networks too (with smart, there are a lot of floating IP associated, the
flat type DHCP agent offers public IP addresses too)
My question is: Am I able to change the segmentation type from flat to
vlan somehow? (both on our "flat" and "smart" networking)
Downtime will be OK for this, but reassigning different IP addresses to
VMs is not.
Thanks:
Peter ERDOSI
_______________________________________________
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
We want to change few thing is our production Openstack setup around
networking. (it's Mitaka now, but soon we planning the upgrade)
Just a few idea, how it looks now:
- We use OVS
- Have 2 network nodes (l3 agent, dhcp agent, etc)
- Have multiple compute nodes
- Have independent controllers (with neutron-api)
You can choose from two network type - please ignore the naming now - :
- "Flat"
- "Smart"
The "Smart" dataflow looks like this: backbone -> network nodes (where
the kernel get down the VLAN tag, and add the untagged traffic to
br-smart) OVS/qrouter -> vxlan tunnel -> compute node OVS -> VM
The "Flat" dataflow looks like this: backbone -> compute node (same as
network node, we have bondX.100, bondX.101 etc interfaces) -> OVS bridge
(br-flat) -> OVS - > VM (the network nodes also have interface in our
"flat" networks, since they doing dhcp/metadata, but the gateway is two
HW router with vrrp, not some dynamic qrouter.
Of course, there are firewalls before the VM, but it's irrelevant.
So, our goal is to unconfiugre this bridge mass, and do the vlan
tagging/untagging in OVS.
AFAIK it can work, if I make a trunk port in OVS, and then choose VLAN
type with the right segmentation ID. (I'am possibly okay with the
configuration itself, the next part looks hard)
So, as you can imagine, we have multiple VM-s in "smart" and "flat"
networks too (with smart, there are a lot of floating IP associated, the
flat type DHCP agent offers public IP addresses too)
My question is: Am I able to change the segmentation type from flat to
vlan somehow? (both on our "flat" and "smart" networking)
Downtime will be OK for this, but reassigning different IP addresses to
VMs is not.
Thanks:
Peter ERDOSI
_______________________________________________
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack