Meshing and XRRP

[H-J]

Hello,
we have some vlans (vlan 1..8) and a couple of servers (hp dl360) with 2 NICs. Each vlan is on another location.
Between the locations we have a limited amount of fibre lines.
The vlans are routed by the Default-GW 53xx (A) on the central site. For router redundancy we use a second 53xx (B) with xrrp configured on both sides (see attachment).
Til now, all servers in vlan 1 are directly connected with the 53xx on central site (A). But the 53xx (A) is a single point of failure for server access.
Now we want to use meshing to increase bandwidth and redundancy. Because meshing and routing isn't allowed on the same switch and the limitation of fibre lines between the central site and the location of switch D, we got two 34xx (E and F). There we want to split the NICs of each server and build a meshing domain including the 34xx switches C and D and the non-routing 53xx switches E and F.
So we need the connection a) for the mesh. But this seems to be a shortcut in the xrrp- functionality.
Is there any problem in this constellation and can you tell me best practices ?
Thanks for your help and let me know if I am unclear...

Best regards,
H-J

[Matt Hobbs]

Hi H-J,

If you're looking at using Meshing, I would actually use more links between the switches. In your diagram, I'd put another link between C & F, and D & E.

I'd also put a link directly between A & B.

Spanning-tree needs to be enabled on all the switches in the network. You need to set the spanning-tree priority higher on the core XRRP switches. (You probably already have this set).

Finally every switch needs to be configured with all of the VLANs in use on the network and have those VLANs tagged on the uplink ports back to the core. If you don't do this, in a failover situation you may notice that some VLANs no longer have full connectivity.

Hope this helps.

Matt

[H-J]

Hi Matt,

thank you for your response.

That's O.K. with the links between C & F and D & E (see actualized attachment).

But I think, that the link between A & B will never go up. Because XRRP is configured at A & B, so B is permanently in Fail- Back- mode. And if one of these switches go down then the link between A & B goes down, too. Or is it wrong ?

Are there any problems to expect with load-balancing on the servers or is there a better constellation to increase bandwidth and redundancy ?

Best regards,
H-J

[Matt Hobbs]

Just looking closer at this, do you actually have spanning-tree enabled throughout at the moment? I'm guessing that you don't.

As soon as spanning-tree is enabled it's going to block a lot of the links that are currently in use which is why you would use the A-B link and enable all VLANs to be tagged on every link to every switch.

It really needs to be planned out on paper properly. I would recommend you get in touch with a ProCurve technical consultant in your region if possible.

For the server load balancing you would probably use Transmit Load Balancing.