Introduction
We have seen scenarios where Cisco DCM (Digital Content Manager) D9902 is connected to a L2 switch and configured to receive unicast video streams. However after streaming about 5 minutes, the same LAN observed flooding on the switch which caused outage in customer network. Later on it was identified that the switchport connected to the DCM aged out the mac-address table which in turn caused the flooding since the destination mac-address was unknown to switch.
Theory
UDP (User Datagram Protocol) flooding is common issue in unidirectional scenario. The ARP (Address Resolution Protocol) cache timer (default 4 hrs) in router/L3 switches is always higher than mac-address age timeout (default 5 min). Means always there is a chance of wiping out the mac address information from switch if there is no response from destined device. Increasing the value of mac table aging time is also not recommended since it could create significant load on the switch and potentially run out of resources.
Solution/Recommendation
There are different ways we can follow to overcome this situation.
- Most reliable and easy solution is to create a dummy multicast in the DCM. In this case DCM will send IGMP (Internet Group Management Protocol) “membership report” to the switch and switch will start polling the DCM periodically by sending IGMP “membership query” which will refresh the mac-address table in switch.
- In the switch, decrease the value of ARP cache timer less than/close to mac table aging timer. This will cause the ARP packets to be broadcast and relearning mac-address must occur before L2 entry ages out.
- Configure static mac-address entry in the switch which will be persistent even after reboot and there won’t be any issue of timeout.
More information about UDP flooding will be available at http://www.cisco.com/c/en/us/support/docs/switches/catalyst-6500-series-switches/71079-arp-cam-table...
