ACI Fabric Clears Endpoints from EPG After a Few Hours – Causing External Traffic Drop
Description:
Hello everyone,
I'm facing an issue in my Cisco ACI fabric related to endpoint persistence. I have an EPG associated with VLAN 1485, and a bridge domain also mapped to VLAN 1485. The EPG is correctly configured, and the VMs connected to this EPG can successfully ping the gateway of the bridge domain. As expected, the endpoints are initially learned on the corresponding leaf switches, and communication within the fabric works fine.
However, after a few hours of inactivity, all learned endpoints associated with this EPG are cleared from the fabric. When this happens, the APIC shows no endpoint clients under the EPG, and consequently, any traffic initiated from outside the fabric (e.g., from an L3Out or external source) towards these VMs fails, as ACI no longer knows how to forward the traffic due to the absence of the endpoint information.
This appears to be a classic case of endpoint aging, but I would like to understand:
-
Why ACI is not relearning or retaining the endpoints once they are aged out, especially when external traffic is trying to reach them?
-
What settings should I review in the Bridge Domain (e.g., ARP flooding, unknown unicast flooding, endpoint retention policy)?
-
Are there best practices to keep endpoints active (e.g., GARP, static bindings) when dealing with idle or low-traffic VMs?
-
Is there any way to allow ACI to forward traffic to these VMs even if the endpoint is not currently present in the endpoint table?
I'd appreciate any guidance or suggestions on how to resolve or work around this issue.
Thanks in advance!
