How do APICs in a cluster communicate with each other (in particular DB exchange)?

[rund]

Hi,

i can't find any usefull details on this.

How do APICs in a cluster communicate with each other over multiple leaves? Somehow they must synchronize their database (sharding), but how is that done in detail (default operation)? Are they using a VXLAN overlay network? Is the Infra- or the Management-Tenant used? Is the traffic sent somewhere in-band?

Thanks a lot!

[Manuel Velasco · Accepted solution]

Hi Stefan,

The APICs communicate via the infra vlan using the infra network. They get assign an IP from the TEP pool that is configure during the initial setup.    The APICs would get assigned the first IPs of this pool. 

If you want to know know the IPs for the controllers you can ssh to one of the APICs and the the “show controller’ command.   

If you want to know the IPs from the nodes, from the apic you can run the “acidiag fnveread” command.

From the leaf nodes you can reach the ip of the controller using the overlay-1 VRF. You can ping the controller IPs using the “iping -V overlay-1 <ip address of controller>

Please let me know if this is what you were looking for.

[rund]

Very helpful, indeed. Exactly what I was searching for. Thanks!