From: Roche-Kelly, Edmund B. (Edmund.B.Roche-Kelly@xxxxxxx)
Date: Mon Jun 11 2001 - 09:55:57 GMT-3
You can use etherchannel between Access A and Access B. Preferably on a 6500
so
you can channel ports on different modules. If you use the ports on the
supervisor module, with 1/1 and 2/1 going to distribution, and 1/2 and 2/2
going
to the other access layer switch, a scenario where both 1/2 and 2/2 are down
and the
links to the distribution are still up becomes unlikely. Avoiding spanning
tree does
bring up a whole new set of issue, but there's really nothing uglier than a
spanning tree
loop.
Ed
-----Original Message-----
From: Yves Fauser [mailto:Yves@Fauser.de]
Sent: Saturday, June 09, 2001 6:22 AM
To: Roche-Kelly, Edmund B.
Cc: 'Williams, Glenn'; Ccielab (E-mail)
Subject: Re: Spanning Tree & HSRP
Hi,
I was asking myself what happen4s in the so called U shape if the link
between
ACCESS A and ACCESS B goes down. Then you would have the same situation as
you
have in your second example. Both Interfaces on the RSM would stay up, and
both
would get active in HSRP. In this Situation the routing protocol would no
switch
the path over from one RSM (Distribution switch) to the other, and would
still
load balance between them if there are equal cost path.
Am I right, and if not, where is thing that I4m missing ?
Yves
P.S. I will not be able to read my mails until Tuesday06/12/01, so if there
are
further discussions, sorry that I can be part.
"Roche-Kelly, Edmund B." wrote:
> This loop free design is the latest fad in Cisco design circles, mainly
due
> to 4 years of spanning tree loops breaking networks. The idea is to keep
> each vlan
> on only one access layer switch. This gives a V shape instead of the
> traditional
> triangle.
>
> DISTRIBUTION A----vlan 10----ACCESS A----vlan 10----DISTRIBUTION B
>
> If a vlan needs to be on two access layer switches, you use a U shape.
>
> DISTRIBUTION A----vlan 10----ACCESS A===ether channel===ACCESS B----vlan
> 10----DISTRIBUTION B
>
> You can trunk multiple vlans between the Access layer and the
distribution,
> but the
> point is that vlan 10, for example, is on only one link between the
> distribution and
> the access layer. If that link goes down, the vlan interface on the RSM
goes
> down since
> there are no active ports on the switch in the vlan. HSRP works fine.
>
> If you did something like this:
>
> DISTRIBUTION A----vlan 10----ACCESS A----vlan 10----DISTRIBUTION B
> |
> |
> vlan 10
> |
> |
> ACCESS B
>
> Then the vlan interface would not go down on the first RSM if the link to
> ACCESS A broke, because of the active port to ACCESS B and both RSMs would
> be HSRP
> active, with standby router unknown.
>
> Spanning tree is very unfashionable since layer 3 switching means the old
> routing latency
> is gone, and layer 2 recovery times are so much longer than routing
protocol
> recovery times.
>
> Ed
>
> -----Original Message-----
> From: Williams, Glenn [mailto:WILLIAMSG@PANASONIC.COM]
> Sent: Thursday, June 07, 2001 8:51 AM
> To: Ccielab (E-mail)
> Subject: Spanning Tree & HSRP
>
> Hi all. This is an actual design in progress that was sent to Cisco for a
> response. As we await, I wanted to see your comments. From what I
> understand (???) Cisco would like to see switch designs without any loops
in
> the initial design so as not to depend on spanning tree. (However even
> though it may be awkward, don't loops provide for redundancy?) But this
has
> me wondering about a redundant distribution switch design involving HSRP
> with two RSMs inside two 4006s. Below is what I sent to Cisco. What do
ya
> think? BTW I cannot test this yet and I'm not as learned as many of you.
So
> I appreciate your input.
>
> Message to Cisco:
>
> We will be implementing 2 distribution switches with RSMs that will
connect
> to all our access level switches. This will provide fault tolerance. The
> RSMs will run HSRP. In many Cisco designs I noticed that a trunk is
usually
> run between the two switches. I imagined that this is a way for the two
> HSRP routers to maintain communication and as a path in case a port that
> connects to a particular access switch fails or the wire connecting the
> access to distribution switches fail. However I heard that we may be not
> putting in a trunk between the two distribution switches which has posed a
> technical question that I have not been able to get my mind around. I
> understand that by not putting in the trunk we eliminate a possible loop
> that would be a problem if spanning tree were disabled. I also understand
> that we may or may not put in a separate VLAN between the two distribution
> switches for routing updates or some other purpose. (We were thinking of
> turning off any routing protocols and putting in a default route to our
WAN
> 7505 router.)
>
> Here is what I do not understand.
>
> The only path for the two distribution switches to communicate is either
> between the VLAN (if we connect it) between the two switches or the VLAN
> that would go from a distribution switch to an access switch, then back
down
> to the other distribution switch. If I break the wire to one of the
access
> switches, what happens, especially for HSRP?
>
> I am assuming that the hello messages were flowing from one distribution
> HSRP router to the other via the connection through the access switch.
With
> this connection broken the backup HSRP router becomes active. But
wouldn't
> the original active HSRP also stay active since it did not fail, it just
> lost connectivity to one switch in that VLAN. This being the case, when
it
> receives a packet from another subnet, destined for that VLAN that the
> access switch was on won't it drop the packet? This is why I thought we
> need a trunk between the two so there would continue to be a path to the
mac
> address being sought and so the HSRP routers could continue to talk.
>
> Straighten me out.
>
> Thanks
> Glenn
> **Please read:http://www.groupstudy.com/list/posting.html
> **Please read:http://www.groupstudy.com/list/posting.html
**Please read:http://www.groupstudy.com/list/posting.html
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:31:22 GMT-3