From: jsaxe@Crutchfield.com
Date: Tue Apr 22 2003 - 13:56:27 GMT-3
We have a number of 2924XL's around here, and they are perfectly wonderful
switches, but they do not support dynamic trunking or dynamic port
aggregation (EtherChannel). They will *do* trunking and EtherChannel, but
not the "desirable" or "auto" modes of some of the bigger
(Cat4000/5000/6000) or more recent (2950) switches. So to get a 3550 to
trunk to a 2924, you'll have to use "switchport mode trunk" on both ends:
the 2924 is never going to trunk unless you do it, and the 3550 might if it
were to receive a "yes" answer during DTP negotiations, but the 2924 doesn't
even do that much. So you just have to nail up these features on both sides.
Not too much of a limitation, and it's completely documented in Cisco's
documentation... they even have examples of how to set it up with various
equipment on the each end (both IOS-based and CatOS-based).
By the way, the new 2950's are really excellent on the dynamic trunking
front. They are set by default to have "desirable" trunking, which means as
soon as you attach two of them together with a crossover cable they
immediately trunk, and if you've set up a VTP domain on the first one, the
second one immediately joins it and learns all the VLANs, etc. Quite
convenient. But if you attach a plain workstation cable, it takes a few
seconds to determine that it's not a DTP peer on the other end and it switch
to access mode instead. (Note that I'm glossing over the security
implications of negotiated trunking, and I'm assuming nobody
opportunistically plugs in a switch somewhere in my building. In a secure
datacenter, I'd of course lock down all the ports to access-only, so a
customer couldn't leap into trunking mode and walk all over my layer 3
partitioning and access lists.)
And it's always been sort of dangerous to mix spanning-tree portfast mode
with trunking; you might want a non-trunk port to light up and send traffic
immediately for a workstation, but you want a trunking port to go through
listening and learning states to determine the correct spanning-tree
topology for each VLAN. The 2950 again shines here: If you set an interface
to "spanning-tree portfast", it only means portfast if it becomes access
mode; if it negotiates to trunking, it does not skip ahead in the STP
algorithm, unless you also say "spanning-tree portfast trunk" (dangerous).
Plus the 2950 reloads a heck of a lot faster. Man, the 2924 sucks in that
regard.
-- Jeff Saxe
Network Engineer, Crutchfield Corp.
CCIE #9376
-----Original Message-----
From: Ian Stong [mailto:istong@stong.org]
Sent: Friday, April 18, 2003 9:15 AM
To: 'Aldo'; ccielab@groupstudy.com
Subject: RE: Cat 2924XL and cat3550 EMI
As far as I know if you learn everything you can on one 3550 with EMI the
only other things you would need a second switch for is certain switch
functions. Those "functions" being etherchannel, trunking, vtp stuff and so
forth. So for the second switch as long as it supports all the L2 stuff
like the 3550 then you shouldn't have to worry about L3 as you can use a
router for that. Hence a 2924XL should work marvelously as the second
switch.
This archive was generated by hypermail 2.1.4 : Thu May 01 2003 - 13:36:01 GMT-3