RSTP implementation - Do we really need to identify non-edge ports?

[Preston Gallwas]

I was reading: ftp://ftp.hp.com/pub/networking/software/59908853-1004-Adv-Trff--ch05-STP_RSTP.pdf

It talks about identifying non-edge ports and identifying which ports are edge ports so they immediately come online in forwarding mode. We've got hundreds of links between switches, many of which are not standardized for which ones uplink.

What is the simplest way to implement STP on a "flat"(?) network?

Could...it really be as simple as

"spanning-tree protocol-version rstp"
"spanning-tree"
?

[Preston Gallwas]

well...it appears if I want my uplink ports to stay gigabit , then yes, we do need to identify the edge ports, unless I am mistaken (Did some testing here)

Is there a way to easily identify a switch-to-switch link on HP 2600 series 8.53 firmware (other than "show mac" and looking for the port with the most ;-) )

[Ardon]

Hi Preston,

No although we mostly set it to "Non-Edge" when it is indeed a Switch to Switch Link. Per 802.1W IEEE (RSTP) specification. When a port set to "Edge" receives a BPDU, it will loose it's "Edge Port" status.
So in a simple scenario where you only interconnect three Switches in a triangle with 802.1W enabled, leaving the defaults (Edge)will NOT result in a Layer 2 loop as the ports will "hear" the BPDUs on the inter switch link ports and therefore act as normal ports.
If wanted I can look up and send you the pages of interest from the IEEE 802.1W specs.

Regards, Ardon

[Preston Gallwas]

Hmm... so what is the advantage of setting it to a non-edge port? We want the most optimal setup and to follow best practices, not to short ourselves some work. (at the same time, the easier / faster config is welcomed haha)

[Preston Gallwas]

I reread your response -- you said when a port receives a BPDU, it loses its edge port status

our links switch to switch are mostly gigabit (ports 25 or 26 on 2626 or 49 and 50 on 2650's) -- are you saying that it is possible for someone to create a loop and have that port lose its edge status and begin blocking?

I'd want the new link(which is 100mb) to be blocked

(99.9999% of our loops are users plugging in their network cable back to the wall jack which patches to wiring closet)

[Ardon · Accepted solution]

Hi Preston,

Let me answer your latest posting step by step.

I reread your response -- you said when a port receives a BPDU, it loses its edge port status

our links switch to switch are mostly gigabit (ports 25 or 26 on 2626 or 49 and 50 on 2650's) -- are you saying that it is possible for someone to create a loop and have that port lose its edge status and begin blocking?

Ardon>Yes, and you should be glad as otherwise a Layer 2 loop WOULD occur. By looping two ports, the BPDU will get out and return to the Switch. This mechanism helps to identify loops and take them out by blocking a link.

I'd want the new link(which is 100mb) to be blocked
Ardon>I can not tell as I do not know the topology and configs you are using. Most of the time Blocking Links can be manipulated by setting STP Priority on the Switch and therfore the STP Root in the network.


(99.9999% of our loops are users plugging in their network cable back to the wall jack which patches to wiring closet

Regards, Ardon

[Preston Gallwas]

I guess my question is, is RSTP smart enough to block the 100mb port instead of the 1gb port

Secondary question to that is, if a port is blocked, can PCM+ Find ports that are blocked (aka looped) so that we can remove them from the network as we get around to it? (it wont cuase a disruption, but we still prefer to educate our users)

[Manfred Arndt]

Hi Preseton,

Yes, the PCM network map shows STP blocked links as dashed lines.

You can also you a MIB browser to identify all STP blocked ports using the Bridge MIB (RFC 1493). Look for ports where the dot1dStpPortState is blocking(2).

Cheers

[Preston Gallwas]

That, of course, depends on PCM starting to work again (grrrrrrrrrrrr)

[Manfred Arndt]

Preston,

Sorry about your PCM problems :(

Regarding your other question. Yes, RSTP is smart enough to block a lower speed port.
There is a per-port "path-cost", where the lowest value is used to determine which ports are forwarding.

The following is the default path cost when using the auto option (recommended).

10 Mbps - 2,000,000
100 Mbps - 200,000
1Gbps - 20,000

Good luck

[Preston Gallwas]

Whew, thanks

Last question: the command spanning-tree point-to-point-mac force-false

How critical is it to identify the hubs? We have a great number of HP WL420's out there -- as well as just a few hubs, but most of our infrastructure is

[Manfred Arndt]

Preston,

Regarding the point-to-point-mac question:

It is not important to identify the hubs or the WL 420s, as long as you keep them as Edge Ports (the default).

It is recommended to set switch to switch links as "non-edge" to speed up RSTP convergence.

[Preston Gallwas]

Okay. Thank you for all of your assistance!