Rapid STP and learning state

acidsatyr

RSTP superior convergence time comes from port synchronization mechanism and topology change advertisements which do not rely on specific timers (as it happens in 802.1d). But there’s something that confuses me, for example,
It takes 3 x hello interval or 6sec for a switch (SW1) to notice that something’s wrong with its neighbor (SW2), in which case SW1 (port to SW2 was blocking/discarding) moves that port from discarding to learning. Now learning state is 15 sec and is not yet forwarding; but then how is convergence always sub second?
I hope you understand what I'm saying I’m having hard time putting this into words..
thx for help

cisco_trooper

Make sure you are distinguishing between 802.1d and 802.1w.

acidsatyr

Sort of what i'm trying to do here.

What's the role of learning state in RSTP? Isn't port suppose to go directly from discarding to forwarding (say because blocked port recieved superior bpdu), and then send TC notification to neighbors so they could empty their CAM table? Whats the point of learning state if its not forwarding?

cisco_trooper

Look into RSTP Port Roles and Port Types. Then look into how RSTP detects changes in topology.

cowood2676

Trooper is right, I think you may not fully understand the types and states of RSTP. take a look at the article below. After you read it if you still do not understand I will try to help you out.

Rapid transition is the most important feature introduced by 802.1w. The legacy STA passively waited for the network to converge before it turned a port into the forwarding state. The achievement of faster convergence was a matter of tuning the conservative default parameters (forward delay and max_age timers) and often put the stability of the network at stake. The new rapid STP is able to actively confirm that a port can safely transition to the forwarding state without having to rely on any timer configuration.

Understanding Rapid Spanning Tree Protocol (802.1w) [Spanning Tree Protocol] - Cisco Systems

acidsatyr

Thx for helping out guys,
I understand all of that, I read it couple of times, and I’ve experienced convergence time of rstp. What did throw me off is, *if the ports go through process of synch, on wire speed (or rather on exchange rate of bpdus), instead of specific timers, then what is the point of learning state? When does a port actually go through learning state and why? That's the part that baffles me. For example, when root port misses 3 x hello, it momentarily forwards on alternate port, etc, etc. Wheres learning state part in all of this?
When topology change happens switches notify each other with TC hello flag set so they could all flush their CAM. In 802.1d, learning state was there so switches would be ready when port goes into forwarding state
Here is the question from a book that actually got me thinking:
“Assume that a non-root switch (SW1) is blocking on a 802.1q trunk connected to SW2. Both switches are in the same MST region. SW1 ceases to receive Hellos from SW2. What timers have an impact on how long switch SW1 takes to both become the Designated port on that link and reach forwarding state?
a) Hello timer setting on Root.
b) Maxage timer on Root.
c) Forward delay timer on Root.
d) Hello timer setting on SW1.
e) Maxage timer on SW1.
f) Forward delay timer on SW1.
MST uses RSTP so I answered a) – SW1 missed 3 x hello setting advertised by root, and port trough synch goes to forwarding state.
The book says it’s a) AND c).
Thx again for any help, much appreciated.

acidsatyr

:< .........

acidsatyr

Can anyone shed some light on this?
That question is from CCIE R&S 2007.
What's the point of forwarding delay advertised by root?

EdTheLad

You'll find you answer in any book that explains STP. All BPDUs are generated by the root, forwarding delay is configured on the root and propagates to all other switches via the bpdu.Now just read why forwarding delay is used and you'll have the answer.

acidsatyr

EdTheLad wrote: »

You'll find you answer in any book that explains STP. All BPDUs are generated by the root, forwarding delay is configured on the root and propagates to all other switches via the bpdu.Now just read why forwarding delay is used and you'll have the answer.

No thats not true for RSTP, BPDUs are sent by every switch, not just root. Topology change (port moving to forwarding state) triggers TC bit in BPDU so neighbor switches flush their CAM table (if they need to), therefore convergance is not dependant on specific timers.
so again, whats the point of forwarding timer advertised by root (as it stands in the question), and learning state (when it comes to topology change)?????????????
THIS IS RSTP, NOT 802.1D

kpjungle

acidsatyr wrote: »

No thats not true for RSTP, BPDUs are sent by every switch, not just root. Topology change (port moving to forwarding state) triggers TC bit in BPDU so neighbor switches flush their CAM table (if they need to), therefore convergance is not dependant on specific timers.
so again, whats the point of forwarding timer advertised by root (as it stands in the question), and learning state (when it comes to topology change)?????????????
THIS IS RSTP, NOT 802.1D

The subsecond failover will in most cases be because of the alternate port role (same as uplinkfast), it will use that port if it looses the rootport. The TC bpdu is only sent if a non-edge port goes into the forwarding state. The only way i see the forward timer being invoked is with compability with 802.1D environments and on edge ports.