Protocols & Communication

Problem

Slow Network Availability

Environment

Modern Cisco managed switches support a protocol called "spanning tree". This protocol is designed to, when the port interface comes up, spend 60-90 seconds scavenging the network through its ports detecting loops in the network. For example if Switch A is connected to Switch B, Switch B to Switch C, Switch C to Switch D and finally Switch D to Switch A. This loop will cause a data loop and effectively reduce the ability of all the switches to 0.

Spanning tree will detect this loop before any data flows and shut down one connection (e.g. the link between Switch B and Switch C). Note that a loop of switches can be used as a valid form of redundancy.

The downside of this that the interface is up so devices expect to be able to talk over the link, however they will get no reply. If the device needs information to start, such as a DeviceMaster, then it will time out and not work properly. It will also mean that ViewX may take longer to connect than normal and servers using NIC teaming may also struggle to become operational in an acceptable time frame.

Fix

The solution is to enable portfast (this is a term used on Cisco routers and may not be generic terminology) on ports of the switch that will not be used by switches. Enabling portfast will mean that when the port interface comes up, it will immediatly start to accept traffic. If the port directly connects to a router, then portfast can also be enabled to ensure any routing protocols update correctly.

Cisco entry on the spanning tree protocol