cancel
Showing results forĀ 
Search instead forĀ 
Did you mean:Ā 
cancel
3481
Views
5
Helpful
3
Replies

RSTP convergence time

Hi

I need to replace switch that is curently the RSTP root.The switch is located in datacenter who has stricts downtime rules.

If i type the  " spanning-tree vlan vlan_id root primary " i can expect what downtime until all the switches converges to the new Root ?

Thanks

2 Accepted Solutions

Accepted Solutions

Peter Paluch
Cisco Employee
Cisco Employee

Adrian,

No guarantees can be given, but if you simply use that command on a switch that is already present in the topology and is intended to become the new root switch, the switchover should take place within a second or two, perhaps even less. A temporary flooding of frames may ensue as the CAMs will be erased due to topology change, but that should cease after a while and should not impair the connectivity.

What is very important to watch for is that the ports towards end stations (hosts, servers, disk arrays, printers, etc.) are configured as PortFast, i.e. edge ports. Otherwise, during the topology change event, they would revert to Designated Discarding state and would need 30 seconds to become Forwarding again. The downtime caused by forgetting to configure edge ports properly can be therefore much, much more pronounced.

Are you also using 802.1Q trunks towards these end stations (i.e. an edge port in trunk mode)?

Best regards,

Peter

View solution in original post

Adrian,

Alright, these trunk ports towards servers must then be configured using the

spanning-tree portfast trunk

command issued directly on these interfaces. Never use this command on a port connected to another switch or Layer2 switching/bridging device. It is appropriate only towards routers, servers or end hosts that for special reasons are placed onto trunk links, i.e. on real edge ports.

Assuming that all your inter-switch connections are configured as trunks (which they generally should be), other access ports are best dealt with by the global configuration level command

spanning-tree portfast default

which makes all access ports to behave as edge ports. Generally, that is what you want. Alternatively, you can configure individual edge access ports using the usual spanning-tree portfast command.

If there are any portions of your network that speak legacy STP, they will be cut off for 30-50 seconds. On RSTP/STP interworking, no rapid convergence is possible without ugly hacks (and those ugly hacks could destabilize your network).

One more question: what is the priority of your current root bridge? The command spanning-tree vlan X root primary can work correctly only if the current root bridge priority uses a priority no lower than 8192. With 4096, the command can refuse to work correctly, and with 0, the command will outright refuse to be accepted. If your current root switch uses a priority of 0 or 4096, it should better be increased to 8192 before replacing the current root switch.

Best regards,

Peter

View solution in original post

3 Replies 3

Peter Paluch
Cisco Employee
Cisco Employee

Adrian,

No guarantees can be given, but if you simply use that command on a switch that is already present in the topology and is intended to become the new root switch, the switchover should take place within a second or two, perhaps even less. A temporary flooding of frames may ensue as the CAMs will be erased due to topology change, but that should cease after a while and should not impair the connectivity.

What is very important to watch for is that the ports towards end stations (hosts, servers, disk arrays, printers, etc.) are configured as PortFast, i.e. edge ports. Otherwise, during the topology change event, they would revert to Designated Discarding state and would need 30 seconds to become Forwarding again. The downtime caused by forgetting to configure edge ports properly can be therefore much, much more pronounced.

Are you also using 802.1Q trunks towards these end stations (i.e. an edge port in trunk mode)?

Best regards,

Peter

Yes,i also have 802.1Q trunks on some ESX servers.

Adrian,

Alright, these trunk ports towards servers must then be configured using the

spanning-tree portfast trunk

command issued directly on these interfaces. Never use this command on a port connected to another switch or Layer2 switching/bridging device. It is appropriate only towards routers, servers or end hosts that for special reasons are placed onto trunk links, i.e. on real edge ports.

Assuming that all your inter-switch connections are configured as trunks (which they generally should be), other access ports are best dealt with by the global configuration level command

spanning-tree portfast default

which makes all access ports to behave as edge ports. Generally, that is what you want. Alternatively, you can configure individual edge access ports using the usual spanning-tree portfast command.

If there are any portions of your network that speak legacy STP, they will be cut off for 30-50 seconds. On RSTP/STP interworking, no rapid convergence is possible without ugly hacks (and those ugly hacks could destabilize your network).

One more question: what is the priority of your current root bridge? The command spanning-tree vlan X root primary can work correctly only if the current root bridge priority uses a priority no lower than 8192. With 4096, the command can refuse to work correctly, and with 0, the command will outright refuse to be accepted. If your current root switch uses a priority of 0 or 4096, it should better be increased to 8192 before replacing the current root switch.

Best regards,

Peter