Configuring first redundant network with extreme networks

[lukas90]

Hello it’s my first time configuring redundant network and I need some help.

My network consists of 1x X620-16X switch and 7x 5320-48P-8XE switches.

Right now X620-16X  acts as default gateway for all vlans and dhcp relay. I have configured 3 LACP groups for esxi hosts. It acts as core switch for all 7x 5320 switches.

I have new switch 5120-24X-4Y-AC-F and I want to configure it as second core switch in case of malfunction of X620-16X.

After adding second routes from 5320 switches to new 5120 it will most likely destroy my network without proper configuring it first. What should I configure on both core switches to avoid disaster.

Can I somehow configure LACP for esxi host to use 1 port each from both of core switches?

What about dhcp relay or default gateway for each vlan? How can I configure same gateway for vlans on botch switches or what is the proper method to do so.

[MTH · Accepted solution]

I'm only passing on some old knowledge that doesn't have anything to do with Extreme switches. If you want the new switch to be your secondary multilayer or core switch, then connect a L2 trunk between the two. Adjust the spanning tree priority to make the first the root bridge and the second the backup by priority. For example, my root bridge has a priority of 4096, the backup is 8192, and all the other managed switches are just at the default, 32768. Since you said the first switch acts as the gateway for all, you have to duplicate that in the secondary by configuring something like VRRP so that there are redundant gateways. A virtual interface is created in both switches say on IP address x.y.z.1, then the physical interface for the VLAN is .2 on one switch and .3 on the other. You have to do this for every VLAN and set a priority so that the active gateway is what you want it to be.

[ØyvindR · Accepted solution]

Hi,

There are several challenges here. Let's start with layer 2 to allow each esxi host to use 1 port from each switch:

My suggestion is to not use LACP, or any kind of interface bundling in the network. Allow ESXi to handle the redundancy. Not only is this much easier, it is also the recommended way from VMWare. This is sometimes called switch independent teaming and will allow switch redundancy. for the ESXi hosts without any configuration in the network (apart from adding the VLANs to the ports). Furthermore, this is the default configuration on ESXi. 

NIC teaming policy: Route based on the originating port ID

The other challenge, default gateway and ip relay is "a bit" more complicated. For layer 3 capable switches, this is usually solved by VRRP, which is its own can of worms, and I wouldn't bet on the 5120 supporting this. It seems quite limited with regard to layer 3 capabilites.

[MTH · Accepted solution]

I'm only passing on some old knowledge that doesn't have anything to do with Extreme switches. If you want the new switch to be your secondary multilayer or core switch, then connect a L2 trunk between the two. Adjust the spanning tree priority to make the first the root bridge and the second the backup by priority. For example, my root bridge has a priority of 4096, the backup is 8192, and all the other managed switches are just at the default, 32768. Since you said the first switch acts as the gateway for all, you have to duplicate that in the secondary by configuring something like VRRP so that there are redundant gateways. A virtual interface is created in both switches say on IP address x.y.z.1, then the physical interface for the VLAN is .2 on one switch and .3 on the other. You have to do this for every VLAN and set a priority so that the active gateway is what you want it to be.