stp

[hakalugi]

stp route determination, or weight/costing question.

we've got fiber connecting far ends of our campus buildings. as a redundant link, incase our fiber is cut by a contruction crew, etc. we have a wifi bridge from the 2 end points.

the majority of our data was using the 1gb fiber (along the right side of the image attached) - but we recently rebooted some of the HPs, and now the primary 'route' used is over our much slower wifi bridge (ie: it's a 99% capacity, when our fiber is almost dark LOL)

in the attached image, the switch at A is a 2626, at B it's a 2650, and at C it's a 2626 - all with the latest firmware as of late October.

can you tell me where to look for how to tell my A & B switches to prefer the fiber route? (STP is not on C, just A and B, so i assume i tell B port 50 costs less than port 1, and same idea for switch A - use port 25 over port 2) - just can't find where to do this via telnet.

thanks.

[hakalugi]

oh, and at far right, the fiber goes 'through' a 2626 with GBIC on 25/26, with about 10 station/node drops on ports 1 - 11.

[Anders_35]

For 2600 series, the syntax for per-port config of stp is

spanning-tree path-cost <1-65535> priority <0-255> mode < norm | fast >

You can adjust path selection by changing the path-cost or priority.

This is described in detail in the advanced traffic management guide.

Anders :)

[hakalugi]

ok, i already had my cost at 20000 on the fiber ports at A& B and (like all the others) 200000 cost on the cat-V port that lead to the wifi bridge

the priority was 8 for both routes, so i changed to 6 on the fiber at both ends, and that seemed to do it.

thanks, i was looking for STP, but because my models weren't mentioned here: ftp://ftp.hp.com/pub/networking/software/6400-5300-4200-3400-AdvTrafficMgmt-Oct2006-59906051-Chap06.pdf i passed over that doc, but it lists what you say :)

[Anders_35]

Chek out the Procurve Networking manuals page:

http://www.hp.com/rnd/support/manuals/

There you'll find the Adv. Traffic Mgmt. guide for the 2600 series.

Anders :)

[DaGuru]

By not including C in the Spanning Tree you may run into other problems. In most cases, you can get away with simply setting the switch priority and let the rest take care of itself. As you will probably discover from reading the guides, you may want to re-think how your tree is being built out.

One area of concern I see is the link you have blocked out in the lower middle area. The connectivity beyond C-24 and B-49. By not including C in the spanning tree you are relying on the switch to bridge BPDUs for you. I have seen this work before, but at the expense of higher CPU load with odd anomalies that were difficult to debug. It is best to use ST on all switches where loops can occur.

[hakalugi]

we used to have a dumb 3com office/stacker there, at 'c'.

when i introduced my last 2626 there, configured as the others (stack member, STP on, my 2 vlans (1 for authorized machines using radius, 9 for all others) nodes on the far end (clients connected to C) couldn't reach servers connected to A.

i tried rebooting C and then A to no avail. so i disabled STP on C - that's when these lost souls got back online, but the wifi link was over utilized.

should i prompt my master to refresh the tree in a manual way when turning stp 'back on' on device C ?

thank you again.

[DaGuru]

Spanning Tree is designed to "heal" or re-converge automatically. So, if designed properly, you should not have to do anything.

I'm having a bit of trouble visualizing your spanning tree environment, so I'll assume that the WIFI link is between A-2 and C-23.

If that is the case, your goal should be to make sure that WIFI link has the highest path cost to your root. I assume that when you say "master" you are referring to your root or the switch with the highest priority (the lowest number).

When you have clients that can't reach their destination, try using SHOW SPAN (abbreviated) to see which ports are blocking/forwarding. There are other commands that can be used to drill down deeper into how STP is operating. With a little experimentation, I'm sure you can find where the trouble is.