From: Carlos G Mendioroz (tron@huapi.ba.ar)
Date: Thu Oct 27 2005 - 12:57:36 GMT-3
I was not forgetting, but adhering to advice :)
So you are using DSWs linked as in the "alternative block design".
Cisco moves from one to the other in different places, and argues that
not having STP convergence (because of no alternative paths in L2) is
the reason for not having DSWs linked.
I'm browsing the BCMSN manual and depending the page you look at, DSWs
are or are not linked :(
Are you running RSTP ? Uplink fast ?
-Carlos
Guyler, Rik @ 27/10/2005 12:36 dixit:
> Don't forget the dedicated link between DSW's. This prevents the DSW's from
> bringing both sides of the HSRP pair to active. Without this link (I prefer
> an Etherchannel for some redundancy) you are right...both sides could go
> active and then you would cause problems for the clients on the ASW's.
> Also, I didn't see in your description but all ASW stacks should have at
> least one link back to each DSW and of course each DSW has at least two
> links into the core on different devices. The number of VLAN's per each ASW
> stack really doesn't matter although there are practical limitations that
> should be observed. Each of my stacks have at least an end-device VLAN and
> a management VLAN. In some cases there may be an additional end-device VLAN
> or two as well but usually no more.
>
> When you approach the design of this area, break things down into
> distribution blocks and consider each block a small network unto itself.
> Beyond that block into the core you should be Layer 3 only so in essence you
> are taking several L2/L3 networks (distribution blocks) and joing them
> through the core. There are also other Layer 2 considerations within the
> distribution block, such as STP optimization that should be addressed in
> coordination with HSRP.
>
> I just replaced all of the Sups in my DSW's (4000/4500) during daytime hours
> without anybody even noticing so the redundant nature of the design is
> solid.
>
> -----Original Message-----
> From: Carlos G Mendioroz [mailto:tron@huapi.ba.ar]
> Sent: Thursday, October 27, 2005 11:07 AM
> To: ccielab@groupstudy.com
> Subject: OT: HSRP in multilayer switching architecture
>
> Hi,
> I'm back again with this topic, "HSRP backtraffic howto" so to say.
>
> While reading (again) the campus network design paper, I see that you can
> use HSRP to have default GW for ASWs (Access switches) to reach the world in
> case of DSW (Distribution switch) failure.
> So basically ASWs do have dedicated IP subnets (usually two for load
> balancing) and then is a DSW goes down, the remaining DSW (they go in
> pairs) takes the whole traffic.
>
> Great.
>
> But what happens if a ASW - DSW link goes down ?
> HSRP still works because DSWs are linked by the failing link, so both
> pretend to be active (isolated I guess) and the one talking to the ASW does
> the job. But what about the traffic going to the ASW ? Both DSW are layer 3
> active on that IP subnet, and unless there is hardware indication of the
> link down at the switch, traffic will be half dropped.
>
> Am I missing something ?
>
> Regards,
> --
> Carlos G Mendioroz <tron@huapi.ba.ar> LW7 EQI Argentina
>
> _______________________________________________________________________
> 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
>
-- Carlos G Mendioroz <tron@huapi.ba.ar> LW7 EQI Argentina
This archive was generated by hypermail 2.1.4 : Sun Nov 06 2005 - 22:00:54 GMT-3