Spanning Tree Protocol saves when Link Aggregation fails

                                                                                         

Traffic storms in the networks, when appear, quickly destroy network services and take equipment down for all the time of the storm's reason detection and troubleshooting. And the troubleshooting, depending on many factors, can take a lot of time while the customers suffer from outage.
Traffic storms can happen even in most unexpected network conditions like Link Aggregation (LAG) setups. Different vendors produce switching equipment with their own understanding of networking standards, and interconnection of devices, made by two different manufacturers, can bring surprises in unexpected place and time. The 802.3ad standard for the aforementioned Link Aggregation requires that all links, included in the specific LAG group (trunk) will be never treated as independent of this group unless they are removed administratively. Links in the LAG trunk can go up and down, but from the link layer (layer 2) perspective they all share the same virtual pipe between two adjucent switching devices. Frames, traversing this pipe in one direction, cannot be forwarded back through any of the physical links comprising the LAG trunk (according to the 802.3ad standard). But here the "specific" manufacturer's implementation is coming and in some point of the switching equipment utilization the traffic storm suddenly appears and disrupt the services. What to do ? Discard the equipment and look for the alternative ? Good suggestion, but you have hardly minutes to restore the disrupted services and after this to take appropriate measures in order to prevent such occurences in the future. So, what is the solution ? The solution is the Spanning Tree protocol (STP). All physical links in the LAG trunk should be apropriately configured to participate in spanning tree algorithm and whenever the switch will recognize its own STP BPDU returned back to one or more of its ports, it will block these ports as redundant ones, thus preventing a bridging loop (and the traffic storm !) and maintaining the network operational conditions.
The links status in this case will be similar to following:

Interface STP Mode State Role

------------------------------------------------------

0/1 Enabled Forwarding Designated

0/2 Enabled Discarding Backup

A change in the STP topology will be alerted by SNMP trap from the switch and a corresponding alarm will appear in the Network Operations Center (NOC), so the condition will be known to the support personnel and corrective action can be applied to the device according to the best practice - or to stay with one link until the next maintenance window and then to reboot device, or to move all the traffic via a redundant set of equipment (if available) and reboot the faulty switch as soon as possible. The proper network design should always include redundancy in different layers (and levels). Thus single point of failure will be always bypassed and proper network operation will be maintained.

About the author: