Two switches for redundancy

[mikiudon]

Hi guys,

I have 3 PROXMOX hosts with a storage network on a 10GB SFP connection to a CISCO Nexus 5500 switch. Backend NAS is a TrueNAS scale also on the same switch with 10GB SFP+ connection.

I am not an expert in CISCO Nexus switches and in network in general so I am posting this question.

Each of the PROXMOX hosts have 1 x SFP+ connection to the switch. Also the TrueNAS has the same 1 x SFP+.

I'm getting another CISCO switch (same one) and trying to see if Linux Bonding from PROXMOX and the TRUENAS would cover redundancy in times of switch failure and hopefully for better storage throughput.

So the idea is to connect the 2nd SFP+ connection to the new switch for both the PROXMOX then the TRUENAS as well. Create a Linux Bond with balance-tlb or balance-alb.

Question...

1. Would this be the right way to do it? I know CISCO switches can do all sort of things but as a base, I want redundancy and better throughput.
2. With 2 seperate CISCO switches mentioned here, can I cross connect them or will this cause issues? I guess one of the switches is now connected to an aggregation switch so it wouldn't be a problem but just want to ask.

Thank you for you comments and sorry if this is not the right forum.

 

[spirit]

Hi,
The best way:

You need to stack/mlag your cisco switches (not sure how nexus works exactly), then you need to create a port-channel across the 2 switches + lacp.

then on proxmox side, create an lacp bond with your 2 interfaces.


(and you can also create a port-channel with 2 links going to your aggreation switch, and create port-channel in your aggreation switch)

Edit: on nexus, it's called : vpc
https://www.firewall.cx/cisco-techn...uration-design-operation-troubleshooting.html

 

[mikiudon]

Hi,
The best way:

You need to stack/mlag your cisco switches (not sure how nexus works exactly), then you need to create a port-channel across the 2 switches + lacp.

then on proxmox side, create an lacp bond with your 2 interfaces.


(and you can also create a port-channel with 2 links going to your aggreation switch, and create port-channel in your aggreation switch)

Edit: on nexus, it's called : vpc
https://www.firewall.cx/cisco-techn...uration-design-operation-troubleshooting.html

Hello thanks for the information. However, knowing I'm not that of an expert for Cisco switches. If I treat the switch as just a 10GB SFP+ switch then would my method also cover redundancy and better throughput? Before I embark, I thought I'd ask people who know better.

Just when switches fail, I don't want to think about LACP and other port channels. Reason being is that I remember we had a LACP based connection and when one port died, the other decided not to work properly so don't want to go through that again.

 

[spirit]

Hello thanks for the information. However, knowing I'm not that of an expert for Cisco switches. If I treat the switch as just a 10GB SFP+ switch then would my method also cover redundancy and better throughput? Before I embark, I thought I'd ask people who know better.

yes, redundancy and throughput.

Just when switches fail, I don't want to think about LACP and other port channels. Reason being is that I remember we had a LACP based connection and when one port died, the other decided not to work properly so don't want to go through that again.

so, you had wrongly configured it

If you don't want to go to lacp path, you can use 2 independant switch, with active-backup bond. (so redudancy, but no better throughput, only 1 link at time)