Re: Basic STP question

From: Petr Lapukhov (petr@internetworkexpert.com)
Date: Thu Jul 17 2008 - 02:07:36 ART

Nate,
We need to clearly separate direct and indirect link failures detection
procedures with classic STP.

If a switch has redundant paths to the root (i.e. has blocking ports), then
as soon as it detects a *root* port failure *directly* (e.g. interface goes
down) it moves the blocked ports into listening state. The port that
receives better root bridge information will eventually be moved to learning
and forwarding state. If there are no alternative paths to the root, the
switch will declare itself a new root and start sending BPDUs accordingly
out of all its ports. Therefore, to recover from a *direct* root link
failure it would take 2xForwadingTime seconds. Note that if a non-blocking
designated port fails, this will just generate topology change notification
and may affect a downstream switch (provided that it can detect this failure
directrly).

As for indirect failures. Classically, the detection of indirect failure is
based on counting missing BPDUs on the root port. The maximum time a port
stores BPDU information is MaxAge-BPDUAge (the age reported in BPDUs
received on the root port). Therefore it takes MaxAge-BPDUAge seconds to
detect that we lost connectivity to the root bridge + 2xForwardingTime to
move best alternative blocking port into forwarding state. Assuming the
worst case it's around MaxAge + 2xForwardingTime delay (50 secodns by
default).

However, as you remember, BackboneFast feature was specifically designed to
speed up indirect link failures detection. As soon as a switch receives an
inferior BPDU on one of it's designated ports (signalig that some other
switch in the network lost connectivity to the root bridge) the switch may
start RLQ procedure looking for a new root port and effectively defeat the
MaxAge penalty. Therefore, with BackboneFast feature, indirect link failure
detection may take approximately the same 2xForwardingTime seconds (maybe a
bit longer).

As for the question above, I believe it was with respect to direct link
failures only. This is why it is enough just to change the ForwardingTime
timer on the root switch in the topology. However, if you are extra
cautinous, you may enable BackboneFast feature on all switches in the
topology. Naturally, the last action only has sense if you have 3 or more
switches. 

-- 
Petr Lapukhov, CCIE #16379 (R&S/Security/SP/Voice)
petr@internetworkexpert.com

Internetwork Expert, Inc.
http://www.InternetworkExpert.com
Toll Free: 877-224-8987
Outside US: 775-826-4344
Online Community: http://www.IEOC.com
CCIE Blog: http://blog.internetworkexpert.com

2008/7/17 Nate Cielieska <ncielieska@gmail.com>:

> Chris,
>
> Based on the fact that a failure constituted the STP topology change.. and
> we are talking about Classical STP (as opposed to RSTP) i would say that
> your maxage comes into play here. Based on failure the way i understand it
> is that STP needs to do a sanity check to make sure that the port going
> down
> is validly going down and adjust its topology accordingly, afterwhich your
> forward delay kicks in.
>
> Contrast to that if you plugged a new port into a switch, there is nothing
> to age out.. so forward-delay x2 and then forwarding would take precedence.
>
> I would be interested to hear on the theory of other master blasters but
> the
> quick way i think about it is:
>
> Existing port into a failure condition -> Maxage applies
> New port coming into the envionrment -> No Maxage countdown.
>
> Regards,
> Nate
>
>
>
> On Wed, Jul 16, 2008 at 5:26 PM, cciestudy <cciestudy@mid-world.net>
> wrote:
>
> > I would read the question to cover only how long it takes for a specific
> > port to go from blocking to forwarding, which would be the listening and
> > learning timers. (2x4sec = 8sec).  So, I would set the forward STP timer
> on
> > the root bridge to 4 sec.
> >
> > -----Original Message-----
> > From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
> > Christopher Copley
> > Sent: Wednesday, July 16, 2008 2:34 PM
> > To: ccielab@groupstudy.com
> > Subject: Basic STP question
> >
> > All,
> >
> > This is one of those basic questions that you dont think about until you
> > have to do it but I am unclear of a requirement and would like everyones
> > thoughts.   STP convergence, I have a question that reads as follows...
> >
> > Configure your Layer 2 network to forward packets in 8 seconds after a
> > failure.
> >
> > My thinking is I would set the Hello time to 1sec and the forward time to
> 4
> > seconds and leave max age alone.  Is this correct? and also should I
> modify
> > the hello time at all?   I set this up in my home lab and it appears to
> > work, but does it meet the requirements of the question?
> >
> >
> > Thanks,
> > Chris
> >
> >
> > _______________________________________________________________________
> > Subscription information may be found at:
> > http://www.groupstudy.com/list/CCIELab.html
> >
> >
> > _______________________________________________________________________
> > Subscription information may be found at:
> > http://www.groupstudy.com/list/CCIELab.html
>
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html

This archive was generated by hypermail 2.1.4 : Mon Aug 04 2008 - 06:11:55 ART