12-10-2018 05:45 PM - edited 03-08-2019 04:47 PM
I am having a hard time getting my PVST+ Rapid Spanning Tree to converge in less than 15 or 20 seconds.
I am using a Cisco IE-5000 as the root bridge and there are loops and subloops involved so I am not saying this is a simple mesh.
The way it goes is:
IE5000A<------------------->IE5000B<----------------------->IE5000C
| | | | |
\/ \/ \/ \/ \/
IE4000A<->IE4000B<->IE4000C IE4000D<->IE4000E<->IE4000F IE4000G<->IE4000H---end
So you can see the loops.
I just configured normal PVST+ Rapid spanning tree. The ports between all the switches are trunked.
The root is the IE5000A. I unplug between IE5000A and IE5000B and I lose packets between a laptop plugged into 5000A and 5000C for 20 seconds. And then I plug the fiber back in and I drop packets again for 20 seconds.
Is there a standard config I should look at or am I missing something. Do I have to play with timers> I thought Rapid would have 2 seconds failover.
James
Solved! Go to Solution.
12-10-2018 11:43 PM
Hi James,
In addition to the suggestions of other friends here, RSTP absolutely requires you to properly configure edge ports using spanning-tree portfast [ trunk ] command. This is because one of the RSTP mechanisms for fast convergence, the Proposal/Agreement, leads to temporary blocking of links to prevent loops, and the unblocking of the links is based on receiving an Agreement BPDU from the downstream switch. However, non-switching devices, such as PCs, laptops, servers, routers, firewalls etc. do not speak RSTP and cannot send back an Agreement - and so, when the temporary blocking hits a port to such a device, this port needs to go through Discarding -> Learning -> Forwarding sequence that takes 30 seconds by default.
I cannot emphasize this enough: In RSTP, all ports that are connected to "end hosts" (or better said, Layer2-terminating device) must be configured with spanning-tree portfast [ trunk ] command (use trunk if the port is a trunk, such as for ports toward servers with VMs, firewalls, or routers on stick). Otherwise, during topology changes, these ports will be blocked and cause outage for 2x forward_delay seconds.
Best regards,
Peter
12-10-2018 06:31 PM
Hi
Have you checked that all the switches are running rapid-pvst? What is the output of "show spanning-tree vlan x" for each switch where vlan x is the vlan the PC is connected to?
Thanks
John
12-10-2018 06:51 PM
12-10-2018 06:58 PM
12-10-2018 11:43 PM
Hi James,
In addition to the suggestions of other friends here, RSTP absolutely requires you to properly configure edge ports using spanning-tree portfast [ trunk ] command. This is because one of the RSTP mechanisms for fast convergence, the Proposal/Agreement, leads to temporary blocking of links to prevent loops, and the unblocking of the links is based on receiving an Agreement BPDU from the downstream switch. However, non-switching devices, such as PCs, laptops, servers, routers, firewalls etc. do not speak RSTP and cannot send back an Agreement - and so, when the temporary blocking hits a port to such a device, this port needs to go through Discarding -> Learning -> Forwarding sequence that takes 30 seconds by default.
I cannot emphasize this enough: In RSTP, all ports that are connected to "end hosts" (or better said, Layer2-terminating device) must be configured with spanning-tree portfast [ trunk ] command (use trunk if the port is a trunk, such as for ports toward servers with VMs, firewalls, or routers on stick). Otherwise, during topology changes, these ports will be blocked and cause outage for 2x forward_delay seconds.
Best regards,
Peter
12-11-2018 03:00 AM
12-11-2018 05:34 AM
12-10-2018 06:44 PM
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide