08-26-2016 04:14 PM - last edited on 03-25-2019 01:21 PM by ciscomoderator
Hi Guys,
I would appreciate if you could please clarify, how data flows between leaf and AVS.
In our environment, we are using AVS in 'local switching' with 'VXLAN' encapsulation. We have configured 'AVS Fabric-wide multicast address' and pool of multicast addresses (one per EPG) addresses.
ESXi host is on UCS B-series.
I have following question with this setup:
1. While configuring VDS we configure pool of VLANs and those VLANs are used to define Port-groups. But with AVS, there is no option to define pool of VLANs while using VXLAN encap. so what VLANs will be assigned to port-groups?
2. what role infrastructure VLAN plan in AVS?
3. what are 'AVS Fabric-wide multicast address' and 'pool of multicast addresses (one per EPG)' address are used for?
4. if you could also explain, how data travels (packet walk) between leaf > FI > AVS. For example if a packet has to go from AVS to an endpoint attached to some other leaf switch
if someone can suggest a good document that cover these topics in details
Thanks in advance
NB
Solved! Go to Solution.
08-26-2016 05:00 PM
Hi Navjyot,
1. In AVS, you will be using the VXLAN to communicate between the AVS and the leafs, so there will be no VLANs configured for each portgroup.
2. The infrastructure VLAN is the only VLAN you will use to communicate between AVS. Basically all traffic between AVS and the leaf will use the infrastructure VLAN as the encapsulation.
3. The AVS fabric-wide multicast address is the multicast address within the ACI fabric to identify this specific instance of AVS. If you have multiple AVS configured, you should configure different multicast for each instance. The pool of multicast addresses (one per EPG) are the multicast addresses used by each EPG for multi-destination packets in that EPG.
4. In general, when a VM sends a packet to AVS, AVS will encapsulate it in VXLAN, then encapsulate with the infrastructure VLAN and send it up the VTEP. On the FI, it will only inspect the outer header, and send it up the uplink with the infrastructure VLAN configured. Once the leaf receives the packet from the infrastructure VLAN, it will decapsulate and check the VNID in the VXLAN header to and forwards to the destination as normal in ACI.
Hope this helps.
Jenny
08-26-2016 05:00 PM
Hi Navjyot,
1. In AVS, you will be using the VXLAN to communicate between the AVS and the leafs, so there will be no VLANs configured for each portgroup.
2. The infrastructure VLAN is the only VLAN you will use to communicate between AVS. Basically all traffic between AVS and the leaf will use the infrastructure VLAN as the encapsulation.
3. The AVS fabric-wide multicast address is the multicast address within the ACI fabric to identify this specific instance of AVS. If you have multiple AVS configured, you should configure different multicast for each instance. The pool of multicast addresses (one per EPG) are the multicast addresses used by each EPG for multi-destination packets in that EPG.
4. In general, when a VM sends a packet to AVS, AVS will encapsulate it in VXLAN, then encapsulate with the infrastructure VLAN and send it up the VTEP. On the FI, it will only inspect the outer header, and send it up the uplink with the infrastructure VLAN configured. Once the leaf receives the packet from the infrastructure VLAN, it will decapsulate and check the VNID in the VXLAN header to and forwards to the destination as normal in ACI.
Hope this helps.
Jenny
08-27-2016 04:39 PM
Hi Jenny,
Everything else makes perfect sense except VXLAN/VNID in point 4. what VNID AVS will pick ? if you could please provide an example.
and also, what i can see from APIC>EPG>Operationl>CLient End-Points> every EPG has been assigned with different VXLAN. But I haven't manually configured any VXLAN. so from where these VXLAN numbers are generated?
Thanks
N B
08-29-2016 11:53 AM
Hi Navjyot,
VNIDs are managed by ACI. We do not allocate any pools for VXLAN. ACI automatically generates and keeps track of all the VNIDs it uses.
The AVS will pick the EPG VNID to encapsulate the traffic as it's the only VNID the portgroup will know.
Thanks,
Jenny
08-30-2016 03:39 AM
Hi Jenny,
Thank you very much for your help!!
Regards,
N B
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide