Xjyuk

I have a nagging Rapid PVST problem on some Nexus 9000 switches. Rapid-PVST keeps recalculating 3 to 5 times an hour. We have in this topology (summarized):

 Edge Router Access Layer
+-------------+ +-------------+
| | Eth1/28 Eth1/54 | |
| Nexus9000_1 +-------------------------+ Nexus9000_2 |
| | Vlan350 | |
+-------------+ dot1q Trunk +-------------+
 |Eth1/45 (dot1q trunk)
 |
 Something_Important

SHOW OUTPUT: Nexus9000_1

Nexus9000_1# sh spanning-tree vlan 350 detail | i from|topology|VLAN
 VLAN0350 is executing the rstp compatible Spanning Tree protocol
 Number of topology changes 1348 last change occurred 0:35:39 ago <---
 from Ethernet1/28 <---
 Times: hold 1, topology change 35, notification 2
 Timers: hello 0, topology change 0, notification 0
 ... Output snipped ...

SHOW OUTPUT: Nexus9000_2

Nexus9000_2# sh spanning-tree vlan 350 detail | i from|topology|VLAN
 VLAN0350 is executing the rstp compatible Spanning Tree protocol
 Number of topology changes 1157 last change occurred 0:35:39 ago <---
 from Ethernet1/54 <---
 Times: hold 1, topology change 35, notification 2
 Timers: hello 0, topology change 0, notification 0
 ... Output snipped ...

BACKGROUND

The reason I found the STP recalculations is because we got so many complaints about the device connected to Nexus9000_2 Eth1/45 having 30-ish second outages over and over again. Configuring Nexus9000_2 Eth1/45 as spanning-tree port type edge trunk made the problem much less visible because STP moves into a forwarding state much faster with that port-type.

I checked and know that the interfaces in this diagram are not flapping.

QUESTION

Each of those switches says it received a topology change event from the other switch. That's not very helpful... and I don't want to band-aid the problem with spanning-tree port type edge trunk.

What is the best way to find the root cause of these STP topology changes using the tools available on Nexus 9000 switches?

Please don't respond with show spanning-tree internal event-history all or other show spanning-tree internal commands without explaining what exactly to look for in those commands.

In my case, I was able to solve the problem by turning on debug spanning-tree rstp interface eth1/54 and debug spanning-tree event interface eth1/54 on Nexus9000_2.

The next time a BPDU triggered a calculation, the debug gave me detailed information on what was happening on the switchport.