HP Networking - How to Troubleshoot Spanning Tree Topology Changes

Information

Excessive spanning-tree topology changes in a network can lead to latency, performance, and unnecessary traffic flooding problems. Ensuring a stable spanning-tree topology is a key foundation to a well running network. This document outlines the main causes of spanning-tree topology changes and covers the applicable switch show commands to diagnose the problem.

Spanning-tree topology changes can be triggered by any of the following:

1. Non edge ports in a spanning-tree domain transitioning from online to offline and vice versa. Non edge ports are ports that connect to other networking devices that typically participate in spanning-tree while edge ports connect to PCs and other devices that should not participate in spanning-tree. All current E Series switches (Excluding switches running Comware) have an auto-edge detection feature which will automatically detect if a port is edge or non edge based on if a BPDU is received on the interface. In a stable network, non edge ports (switch-to-switch) should rarely, if ever transition. If a non edge port is transitioning repeatedly (flapping) however, each transition will end up triggering a spanning-tree topology change. 
2. Spanning-tree BPDUs from other network segments or spanning-tree regions with the appropriate flags set to instigate a topology change. 
3. Switch software. Older versions of software do not have the auto-edge feature and non edge ports have to be manually configured.

Details

It’s best to start troubleshooting spanning tree topology changes on the root bridge of the spanning-tree domain.

• Keep in mind that most of the counter statistics is since switch boot. This will contain historical data. It is strongly advised that the counters should be compared over successive data collections. This gives a timeframe to compare the changes to.
• To check for excessive port transitions (port toggling) the easiest output to gather and review is "show tech statistics" which is near the end of a "show tech all" output. Look for interfaces that show a high number of port transitions and then determine if switch believes the interface is edge or non edge which can be seen in a "show spanning-tree" output:
  
• If the switch is not properly auto detecting edge or non edge status, a port can administratively be forced to edge or non edge.
• A switch port can be configured as admin-edge. If it receives a BPDU it will change state away from edge. The command "show spanning-tree detail" will help verify if the port is in fact behaving as an edge port. Look for the value of "OperEdgePort". It should be Yes.
• If edge and non edge ports are properly detected by the root bridge and none of the non edge ports are excessively transitioning then a "show spanning-tree debug-counters ports all instance 0" will show on what interfaces the root bridge is receiving topology change information on. Optionally a "show spanning-tree details" output provides similar information however the debug-counters provide a date and time stamp for the different counters. Also note that the "show spanning-tree debug-counters"... command can be ran for specific interfaces and specific instances in the case of MSTP. In a single spanning tree region (STP, RSTP, or MSTP with no other instanced defined) the instance ID will be "0": 
  
• Once a particular interface has been identified as receiving excessive topology changes, the above steps can then be repeated on the attached neighboring device. 

• In certain cases the neighboring unit may be part of a different network segment or not locally manageable. In this case BPDU (or PVST if the neighbor is a Cisco) filtering can be applied to the interface to prevent the spanning-tree information being received on the interface from affecting the topology of the local spanning tree region. Be advised that applying BPDU filtering to an interface removes it from spanning tree and the port will always transition to a forwarding state, since it discard BPDUs bi-directionally.
• Always check the software release notes for the switches in question as there have been fixes implemented for spanning tree on all platforms.