RSTP - Message Age - BPDU (Count to Infinity)

[Juan Urti]

Hi!

In RSTP, each switch create his own bpdu, but in STP the BPDU only is created by the Root Bridge (RB).

In STP, the message age, tell us about the time when the RB has created the BPDU.

In RSTP, what does message age do?  because every switch has his own bpdu.

Like some web pages, in RSTP Count to Infinity, the message age is the only mechanism to avoid a loop.

Thanks

Juan

Sorry by mi english.

[Rolf Fischer]

Hi Juan,

my understandig is that (independent of STP version) the Port Age Timer is set to MaxAge - MsgAge.

Let's say a BPDU with default MaxAge (20sec) and MsgAge=2sec is received on a port. The Age Timer of that port then will expire after 18sec. After 18sec, if no more BPDUs are received, the BPDU information is deleted.

I pasted a debug capture for demonstration here:

http://pastebin.com/fE94pNTa

This is done with pvstp+, unfortunately I can't test it with rapid-pvstp+ at the moment.

In addition, the MsgAge of a recieved BPDU has to be greater than MaxAge. This is a comparable mechanism to TTL in IP.

Taken from IEEE802.1D-2004:

"The RSTP Port Information state machine (...) checks that the BPDU’s Message Age is less its Max

Age, and if not, will immediately age out the received information."

"To ensure that old information does not endlessly circulate through redundant paths in the network and

prevent propagation of new information, each Configuration Message includes a message age and a

maximum age. The message age is incremented on receipt, and the information discarded if it exceeds the

maximum. Thus the number of Bridges the information can traverse is limited."

This document gives a good explanation:

http://www.cisco.com/en/US/tech/tk389/tk621/technologies_tech_note09186a0080094954.shtml#stp_timers

HTH

Rolf

[Peter Paluch]

Juan and Rolf,

my understandig is that (independent of STP version) the Port Age Timer is set to MaxAge - MsgAge.

I respectfully disagree. The RSTP is different in this aspect.

The Port Age Timer in legacy STP was created for the purpose of taking into account the processing delay of BPDUs as they traverse the network from the root bridge. However, there is no such delay in RSTP. You do not wait for the BPDU to be originated by the root bridge and traverse all the way through the intermediary switches to reach you. Hence, the Port Age Timer has no meaning in RSTP. Instead, RSTP simply ages out BPDUs received on a port after missing them for three consecutive times (3xHello). In other words, aging out the BPDUs in RSTP is not progressive based on increasing distance from the root bridge anymore.

This different approach has changed the resulting meaning of Maximum Age and Message Age fields in RSTP. The Maximum Age effectively works as the maximum allowed diameter of the network while the Message Age is a hop count field. The Message Age is incremented by each switch when the BPDU of its "upstream" switch is received on its Root port. If the Maximum Age and Message Age are equal, the BPDU is dropped. This is also expressed in the quotation from the 802.1D-2004 you have posted yourself.

Please note that the 802.1D-2004 does not describe STP anymore. The legacy STP has been dropped from the 802.1D-2004 and newer versions, and it codifies RSTP instead. The last 802.1D revision that contained the legacy STP description was 802.1D-1998.

Best regards,

Peter

[Juan Urti]

I have agree with both answers (thanks a lot of).

Please check this document, in the part of Count to Infinity.

http://blog.ine.com/wp-content/uploads/2010/04/understanding-stp-rstp-convergence.pdf

Like this, when a switch will progagate bad information about Root Bridge, the other switches will accept the bad BPDU.

When one switch redirect this info, the Message Age from his own BPDU will be 20Seg, and so on, and then the bad info never dies, because every switch generate his own BPDU with ner Message Age.

I´m very confused, about how RSTP solve the Count to Infinity.

Thanks

[Peter Paluch]

Juan,

Like this, when a switch will progagate bad information about Root Bridge, the other switches will accept the bad BPDU.
When  one switch redirect this info, the Message Age from his own BPDU will  be 20Seg, and so on, and then the bad info never dies, because every  switch generate his own BPDU with ner Message Age.

In order for the bad info to persist in the network, it would need to be continuously injected into it. Imagine a similar scenario where you have a Root Bridge and it suddenly goes off. Does the information about this dead Root Bridge remain in the network forever? Absolutely not. Because the switches that were directly connected to the now-dead Root Bridge haven't heard from him for 3xHello interval, they age out all existing information about it and start considering themselves to be Root Bridges, leading to new Root Bridge election. This change propagates in a similar fashion through the network. Remember: the Message Age set to 20 seconds does not mean that the BPDU must expire after 20 seconds. This timer does not really exist in RSTP at all. A BPDU can still expire much sooner because of other events, and in fact, in RSTP, it always expires only due to other events.

RSTP has an additional feature that allows for rapid BPDU ageout: imagine a switch receiving BPDUs on its port (Root port, Alternate port). These BPDUs are obviously sent by a better Designated bridge on the segment. If this designated bridge suddenly starts sending inferior BPDUs, they would normally be ignored until the better received BPDU expires. However, on point-to-point links in RSTP, an inferior BPDU from the Designated bridge immediately replaces the previous superior BPDU stored on the port.

I am not sure if this helps - I understand that this topic is quite convoluted. If there are still questions, please feel welcome to ask. And perhaps it would be helpful if you suggested a simple topology and demonstrated your doubts on that topology.

Best regards,

Peter

[Juan Urti]

Thanks peter and rolf for the time.

I´m trying to understand first, count to infinity problem, and second, how does the max age/message age help to RSTP topology, to solve this problem.

In a ring topology, a switch with one Root Port RP and one Alternate Port ALT, receives a inferior BPDU for his RP. Because the switch has stored another (better) BPDU in his ALT, the switch will change his RP, and inform it  to his old RP (now DP).

This situation has called count to  infinity. Below the link:

http://www.cs.rice.edu/~eugeneng/papers/TON-Ethernet.pdf

           

Like this, this problem will be solved only by this:

"The stale topology information about bridge 1 continues to go

around the cycle in a count to infinity until either it reaches its

MaxAge, Message Age  or it gets caught and replaced by the fresh topology

information."

Why does Max Age help us to solve this problem? Isn´t Max Age renewed every time that a switch redirect the stale info about the old RB?

Sorry by mi english, i´m a disaster.

E.x.: Ring is cut between 5 and 6. SW6 has the BP.

SW 1 is RB, and after that is going down.

SW 2 haven´t BP and then it´s declares RB and sends bpdu to 3 and 7 (starting the race condition).

Later...SW6 receives a inferior BPDU in his RP and it´s changes his RP to his BP.

SW6 sends bad info about the RB to SW7, starting the count to infinity and creating a loop temporaly.

Like this document, Max Age and Message Age solve it 30 seconds later, but i don´t understand why.

I know this:

Max Age, time that a port store a BPDU.

Message Age, time that a bpdu has been send by the switch. In STP, every switch add 1; in RSTP?

Max Age = Message Age -> the bpdu will be dropped.

[Peter Paluch]

Juan,

I will need to go carefully over the PDF paper you have linked in your last thread. It may contain some information I was myself unaware of. However, that will take me a day or two. I will try to respond after I have made myself acquainted with the research in that paper.

Best regards,

Peter

[Juan Urti]

Thanks Peter!!! this problem is very difficult.

Sorry by the link.

Is the Message Age of the BPDU renewed when one switch receive a BPDU, before send it to a designated port? Because the paper, says that que loop is finishing when the message age of the stale bpdu finish.

I´m very confused.

Thanks!

[Juan Urti]

Hi Peter!!! i have probed that the message age will be increment in every switch, and then the bad bpdu will be age out.

I have put the explanation here.

https://supportforums.cisco.com/thread/2191165

Thanks!!!!!!