Wired Intelligent Edge

I'm running the CX switch simulator in GNS3 to test some network changes before I deploy them in production.  I'm starting off small in GNS3 with a flat, Layer 2 network (everything is on vlan 1 in access mode).  I have three switches connected in a ring with a Linux vm connected to each one of them.  I have Spanning Tree configured on all three switches (I've tried both MSTP and RPVST with the same results).  When I run show spanning-tree the output looks like it should.  All of the switches agree on which one is the root and exactly one of the switches has one of it's connections set as 'Alternate/Blocking' which should take care of the loop.  All three switches have the following config

ssh server vrf mgmtvlan 1spanning-tree mode rpvstspanning-treespanning-tree vlan 1interface mgmt    no shutdown    ip dhcpinterface 1/1/1    no shutdown    no routing    vlan access 1interface 1/1/2    no shutdown    no routing    vlan access 1interface 1/1/3    no shutdown    no routing    vlan access 1interface 1/1/4    no shutdown    no routing    vlan access 1!https-server vrf mgmt

I can shutdown links in the loop and Spanning Tree enables the blocking link to maintain connectivity exactly as I would expect it to do.  So far everything is working perfectly.

The issue is that if I start a packet capture in GNS3 on any of the links in the loop (even if it's the link that Spanning Tree is blocking) when I try to ping from one of the vm's to another one I get an ARP broadcast storm and the only way I've found to stop the storm is to manually shut down one of the switch ports.  Not all Ethernet broadcast cause a storm, however.  If I ping the broadcast address of one of the vm's it doesn't issue an ARP (since it's sending to the broadcast MAC address) and those broadcast pings don't cause a storm.   

I really feel like this is a GNS3 issue since I have basically the same setup in my production network and it seems to not be an issue there.  Has anyone else experienced this behavior or does anyone have ideas of what to do about it?  It's not really usable for me as it is.

Here's a diagram of my network.  Sorry, I forgot to include that in my initial post

I'm running the CX switch simulator in GNS3 to test some network changes before I deploy them in production.  I'm starting off small in GNS3 with a flat, Layer 2 network (everything is on vlan 1 in access mode).  I have three switches connected in a ring with a Linux vm connected to each one of them.  I have Spanning Tree configured on all three switches (I've tried both MSTP and RPVST with the same results).  When I run show spanning-tree the output looks like it should.  All of the switches agree on which one is the root and exactly one of the switches has one of it's connections set as 'Alternate/Blocking' which should take care of the loop.  All three switches have the following config

ssh server vrf mgmtvlan 1spanning-tree mode rpvstspanning-treespanning-tree vlan 1interface mgmt    no shutdown    ip dhcpinterface 1/1/1    no shutdown    no routing    vlan access 1interface 1/1/2    no shutdown    no routing    vlan access 1interface 1/1/3    no shutdown    no routing    vlan access 1interface 1/1/4    no shutdown    no routing    vlan access 1!https-server vrf mgmt

I can shutdown links in the loop and Spanning Tree enables the blocking link to maintain connectivity exactly as I would expect it to do.  So far everything is working perfectly.

The issue is that if I start a packet capture in GNS3 on any of the links in the loop (even if it's the link that Spanning Tree is blocking) when I try to ping from one of the vm's to another one I get an ARP broadcast storm and the only way I've found to stop the storm is to manually shut down one of the switch ports.  Not all Ethernet broadcast cause a storm, however.  If I ping the broadcast address of one of the vm's it doesn't issue an ARP (since it's sending to the broadcast MAC address) and those broadcast pings don't cause a storm.   

I really feel like this is a GNS3 issue since I have basically the same setup in my production network and it seems to not be an issue there.  Has anyone else experienced this behavior or does anyone have ideas of what to do about it?  It's not really usable for me as it is.