Comware

Hopefully a simple question:

Why do our ProCurve switches send out RSTP BPDU's on edge-ports every 2 seconds?

I know that a bridge should send BDPU's every <HELLO-TIME>, but why doesn't it send to other switches only?

</HELLO-TIME>

With some of the newer firmware releases on the newer products, you can use the BPDU Filtering feature now to stop this - however it's not really recommended as there is always the potential for a supposed edge-port to all of a sudden not be an edge port when one of your users decides to bring in that switch from home and create a loop on that bringing your network down.

As long as spanning-tree is enabled throughout your network on all ports, then it should prevent most of those types of problems from occurring.

I guess another answer to your question is that it sends BPDU's to all devices because it simply doesn't know for sure what ports are connected to other switches. It's better for it to be safe than sorry.

Thanks, Matt.
That's what I expected to hear, but not really what I wanted :)

It's no big deal, so I'm leaving it here.

Anders :)

Hi Anders,

I'm intrigued as to what you wanted to hear...

Instead of spanning-tree protecting your edge-ports, there is a new feature that's starting to become available on the new switches 'loop-protection'. So you could use the BPDU filter and loop-protection simultaneously to stop spanning-tree packets but still be protected from accidental loops.

My basic understanding of how the loop-proteciton feature works is that it sends multicast frames, if the frame is received back on the same switch, the port is disabled.

This feature is useful because some switches 'eat' BPDU's, so in the scenario I mentioned earlier, even if that pesky end-user brought in their own switch and created a loop, if the BPDU's are getting eaten then the port is not going to be blocked and the loop is going to take down your network.

Either way, if you replace one feature with the other you're still going to have some extra traffic going out that port!

Matt

I'm reaching a little beyond my level of (in)competence here, but my thought was that a port doesn't automagically change to a switch connection.
Thus, a port that was established as an edge port won't change unless it also looses link.

If that argument holds, it should only be necessary to send BPDUs some short time after link establishment.

That would save a few million packets a day, in our tiny network alone.

I see what you mean. With the edge-port status which defaults to on for RSTP, it can automatically change to off if it detects another switch there. I'm guessing this only happens when it detects BPDU's from another switch - therefore if the other switch was not running spanning-tree it would never know about it and continue to run as edge-port yes. It's in this situation that you really need spanning-tree to keep sending out BPDU's on all ports if a loop was created on this other switch.

I'm also at the limit of my expertise and I'm really just saying what makes sense to me (these may not be the reasons for it at all).

Well, it makes sense to me too now.
I guess it stretches the reach of STP (as I understand it) a bit, since it involves loop detection on/through a non-STP switch.

I always figured that sending BPDUs was only meaningful when there is another STP bridge at the other end.

Thanks for enlightening me :)

Anders