[f-nsp] One port depending on another port.

[Drew Weaver]

-----Original Message-----
From: foundry-nsp-bounces at puck.nether.net [mailto:foundry-nsp-bounces at puck.nether.net] On Behalf Of Lee Pedder
Sent: Wednesday, August 10, 2011 3:36 PM
To: Nick Hilliard
Cc: foundry-nsp at puck.nether.net
Subject: Re: [f-nsp] One port depending on another port.

On 10 August 2011 19:55, Nick Hilliard <nick at foobar.org> wrote:
> On 10/08/2011 16:50, Eric Helm wrote:
>> The FastIron platform has something called Protected Link Groups. Sadly,
>> it doesn't appear to be supported on the TurboIron platform yet.
>
> would be interesting to find out if this feature will be supported on the
> next software release for the TIX24, which is reputed to be based on the
> main-line netiron software train.
>


I don't think this is going to meet Drew's requirements. Protected
link groups are simply a way of grouping a number of links and only
having one of them logically active. All links will physically
negotiate and come up, but logically the standby links are treated by
the switch as if they are down, and traffic will not be forwarded out
of them. There is no mechanism to say if link A goes down, then also
physically bring down ports B and C. Unless I am mistaken this feature
doesn't exist on any of the Brocade product lines.

I have played with protected link groups in the past as an alternative
to STP on FESX. It works well in a dual uplink scenario, but is all or
nothing as it cannot be configured per-VLAN.

>

Lee is correct, basically there is a longhaul link between two TI24x switches, if that long haul link goes down I want to also shut down the local interfaces until the long haul link is restored. 

It's Layer3 -> TI24x (Layer2) -> Wavelength -> Ti24x (Layer2) -> Layer3.

It is not a gigantic problem because of the routing protocols on the Layer3 devices on the far ends but it seemed cleaner to me to have it work this way. (and it seems like a feature that would be available in most platforms without too much trouble).

Similar to how most C|D/WDM systems work where the end link is not active unless all links upon the path are available. (although obviously way less complicated.)

So basically just a dependency group where you say interface x and y depend on interface z, so if interface z goes down x and y also go down.

Seems like a pretty simple thing to do.