STP bites !

[steven-dunn]

In a data centre which uses VTP trunks to connect user VLANS to the same core switches that connect server and network VLANs, if a switch is added to suppliment a user VLAN it will affect a trunk and hence spanning tree will run on all VLANS. Can I employ prunning to remove the threat of STP blocking all Server ports when a trunk carrying all (but uncessessarily) VLANs is modyfied ? Will pruning stop a VLAN from being presented to certain areas of a STP domain ? We have noted in a lab that any modyfication to one VLAN can halt all operation on all other VLANs for up to 50 seconds.

Do Cisco offer anything better/faster than PVST+ ?

any thoughts ?

[mfaust]

STP doesn't bite, but if it is not properly configured it will nibble away at you. STP allows you to configure redundency without loops. That is the basic function of STP. If root and root secondary switches are properly configured, blocked ports will allways occur at the access layer. Your servers should be at the core layer. If you don't have any loops, STP is not necessary but be very careful here. A mistake will bring your entire VLAN down. VLAN trunking is also a good thing when properly configured. The reason for VLAN trunking is to save ports (more than one VLAN allowed on a port). VLAN pruning will save bandwidth on trunks by removing unnecessary traffic to certain switches. If a VLAN is not in use on a switch, traffic to that VLAN will be pruned (not transmitted). As for your lab test; I can't be sure what it is you are doing, but there should not be global issues when making changes (what are the changes?) to a VLAN. Per VLAN Spanning Tree should insulate VLANs from these issues. Maybe I don't understand your lab test. Try reading http://www.cisco.com/warp/public/473/103.html. Good luck!

[steven-dunn]

Thanks for your help. Please allow me to clarify:

switch A and B are core and are etherchanneled. Switch C has a trunk to both core boxes. Switch C has two user vlans, 100 & 101. PVST allows blocking at the access layer for Vlan 100 on trunk AC and blocking on trunk BC for 101. This allows a primative form of load distribution accross both trunks. The same happens for switch D with server VLAN 102 and 103. and switch E for level 2 users on VLANs 104 and 105, etc, etc,etc. As all switches are in the same VTP domain, if trunk AC is broken to insert another switch to increase port count, all VLANs re-run STP and hence the holw domain could drop. If i can prune non essential VLANs from certain locations in the domain, can i reduce the risk of STP ?

[mfaust]

You are correct. If you break the AC link to add a switch, STP has to discover the break and recalculate. Try this to speed convergence: Set STP root and root secondary on the two core switches. On the access layer switch with a blocked port, enable uplinkfast. On all switches enable backbone fast if you do not have any 2900xl or 3500 switches (not supported). If you have 2900xl or 3500 switches do not enable backbone fast on any switches. Also, rather than breaking the AC connection to add a switch, would it be possible to add the switch using a port from one of the existing switches? In other words; leave your connections as is (AC, BC) and add a trunk CX.

[rfroom]

There is always 802.1w Rapid Spanning-tree available in newer software.