Leigh Anne Chisholm wrote:
>
> It's a symptom of the problem I wrote about earlier in this thread. When a
> MAC becomes active on the network, the computer isn't able to communicate
for
> the first 50 seconds the port detects the end-system is active. The port
> begins in blocking mode, then transitions to listening, then learning.
> Finally, once STP determines that a looped topology hasn't occurred, the
port
> is set to forwarding mode. This creates havoc with any end-system that
> expects to receive over-the-network information within the first 50
seconds.
> IP, IPX, AppleTalk - all face the same issue.
>
Well, this is picking nits, but the STP forwarding delay is only 30 seconds.
The 50 second delay only occurs if the path to root is lost such that the
root BPDU is not heard for maxage (20) seconds. A leaf user port only takes
a 30-second hit.
> The simple solution isn't to kill Spanning Tree on all switches - that's
the
> "I don't understand the problem so I'll do whatever works and create a
bigger
> problem" solution. The real solution is to enable portfast on all switch
> ports that have end-systems directly connected. The caveat to this is to
> ensure none of the end-systems are capable as acting as a bridge,
forwarding
> packets between LAN segments. Enabling portfast essentially disables
> Spanning
> Tree on a port - and Spanning Tree is used to ensure a loop-free
environment.
>
Portfast doesn't disable STP at all. All it does is cause forwarding to
occur without the conservative delay for listening and learning. The port
still listens for BPDUs, will detect a topology loop, and will go into
blocking to break that loop. But I definitely agree with Leigh Anne that
it's a BAD idea to disable STP!
But before STP ever gets a chance to do anything with a port, three other
phases must complete:
1) speed/duplex auto-negotiation -- max of 3 seconds per the standard.
2) negotiation of trunking via DTP
3) negotiation of Etherchannel via PAgP
The last two are typically very fast if the device on the other side
is capable of negotiating. If not, then the retries for each can add
up to as much as 15-20 seconds, depending on platform and code release.
In CatOS, the macro "set port host" disables both of these and enables
STP fast-forward/portfast. You can observe the progress by making the
logging a bit more verbose: "set logging level spantree 6".
And while we're at it, enable bpdu-guard so if someone does back-door
and create a loop, the portfast-enabled port will be disabled.
I'd love to see if that makes the Macs happy.
Marty Adkins Email: [EMAIL PROTECTED]
Mentor Technologies Phone: 240-568-6526
133 National Business Pkwy WWW: http://www.mentortech.com
Annapolis Junction, MD 20701 Cisco CCIE #1289
Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=2949&t=2564
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
