Mailing List Archive

LSP Metrics
For a resilient solution recently I've had two firewalls in two DC's connected to two separate VPLS instances. This is through an XMR core running 5.4E

I've created a TE LSP that doesn't follow the shorted path unlike the primary LSP. I've then set up the second VPLS to specifically use the second LSP.

I only want to use this LSP for this VPLS. I don't want any other traffic using this LSP.

I've set the metric for the LSP to 50000, however other VPLS/VLL instances are still using this path. From the config guide it seems that only BGP and IGPs take note of the metric of the LSP. How is it possible to ensure that a VPLS/VLL will only use this LSP if I specifically request it to do so?

Thanks
Darren
http://www.mellowd.co.uk/ccie
Re: LSP Metrics [ In reply to ]
> For a resilient solution recently I've had two firewalls in two DC's connected
> to two separate VPLS instances. This is through an XMR core running 5.4E
>
> I've created a TE LSP that doesn't follow the shorted path unlike the primary
> LSP. I've then set up the second VPLS to specifically use the second LSP.
>
> I only want to use this LSP for this VPLS. I don't want any other traffic using
> this LSP.
>
> I've set the metric for the LSP to 50000, however other VPLS/VLL instances
> are still using this path. From the config guide it seems that only BGP and
> IGPs take note of the metric of the LSP. How is it possible to ensure that a
> VPLS/VLL will only use this LSP if I specifically request it to do so?
>
> Thanks
> Darren
> http://www.mellowd.co.uk/ccie
>

You can specify an LSP by name in a VPLS by using:

vpls-peer <ip> lsp <lsp>
from within your vpls configuration.

VLLs on the other hand, you'll need a software upgrade (not sure it's available yet)?

_______________________________________________
foundry-nsp mailing list
foundry-nsp@puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp
Re: LSP Metrics [ In reply to ]
Hi Darren,

I don't think there is a good way of keeping other VPLS and VLLs off a specific LSP, but there are some options.

The LSP metric doesn't work for L2VPN traffic, so it's not useful. To keep other VPLS off the LSP, you could configure all of them to use other LSPs. Obviously this comes with a lot of configuration overhead, and it doesn't currently work for VLL (though soon, hopefully).

The only way I know of to do this for VLL (which also works for VPLS), is to set the class of service for every VPLS and VLL. They will never use an LSP with a higher COS than their own configured COS, so by setting the LSP COS appropriately you can probably keep them on the right LSP.

Cheers.
Mitchell

_____

From: Darren O'Connor [mailto:darrenoc@outlook.com]
To: foundry-nsp@puck.nether.net [mailto:foundry-nsp@puck.nether.net]
Sent: Wed, 19 Feb 2014 21:43:06 +1100
Subject: [f-nsp] LSP Metrics


For a resilient solution recently I've had two firewalls in two DC's connected to two separate VPLS instances. This is through an XMR core running 5.4E

I've created a TE LSP that doesn't follow the shorted path unlike the primary LSP. I've then set up the second VPLS to specifically use the second LSP.

I only want to use this LSP for this VPLS. I don't want any other traffic using this LSP.

I've set the metric for the LSP to 50000, however other VPLS/VLL instances are still using this path. From the config guide it seems that only BGP and IGPs take note of the metric of the LSP. How is it possible to ensure that a VPLS/VLL will only use this LSP if I specifically request it to do so?

Thanks
Darren
http://www.mellowd.co.uk/ccie
Re: LSP Metrics [ In reply to ]
Unfortunately that's not feasible. I'm approaching many hundreds of VPLS on many XMR boxes and moving all of the others to another named LSP just won't work.


Ideally Brocade need to ensure L2VPN also takes the metric into account, or be able to tag an LSP somehow so that it's only used if specifically configured to do so.



Thanks
Darren
http://www.mellowd.co.uk/ccie



> From: timoid@timoid.org
> To: foundry-nsp@puck.nether.net
> Date: Wed, 19 Feb 2014 11:13:52 +0000
> Subject: Re: [f-nsp] LSP Metrics
>
> > For a resilient solution recently I've had two firewalls in two DC's connected
> > to two separate VPLS instances. This is through an XMR core running 5.4E
> >
> > I've created a TE LSP that doesn't follow the shorted path unlike the primary
> > LSP. I've then set up the second VPLS to specifically use the second LSP.
> >
> > I only want to use this LSP for this VPLS. I don't want any other traffic using
> > this LSP.
> >
> > I've set the metric for the LSP to 50000, however other VPLS/VLL instances
> > are still using this path. From the config guide it seems that only BGP and
> > IGPs take note of the metric of the LSP. How is it possible to ensure that a
> > VPLS/VLL will only use this LSP if I specifically request it to do so?
> >
> > Thanks
> > Darren
> > http://www.mellowd.co.uk/ccie
> >
>
> You can specify an LSP by name in a VPLS by using:
>
> vpls-peer <ip> lsp <lsp>
> from within your vpls configuration.
>
> VLLs on the other hand, you'll need a software upgrade (not sure it's available yet)?
>
> _______________________________________________
> foundry-nsp mailing list
> foundry-nsp@puck.nether.net
> http://puck.nether.net/mailman/listinfo/foundry-nsp