I'm planning to re-cable a couple of VM hosts that each have 2 x 1G ports and 2 x 10G ports. The switches have a few 10G ports available, but they're somewhat limited. Currently, both VM hosts are connected to the same switch using all their ports for redundancy. I'm considering connecting one 10G link between the two VM hosts and using the other 10G ports for the rack switch, with the 1G ports serving as a failover option. Is this a good approach, or are there any risks I should be aware of? Also, any resources you recommend I check out before making these changes?
2 Answers
How many servers are you running, and are they in a cluster? Do you use shared storage? Typically, it's best to have a dedicated interface for management, a couple for frontend traffic, and some for backend (like storage traffic). Given your setup of 2 x 1G and 2 x 10G ports, I’d suggest something like this:
- Management: 1G
- Frontend: 1G
- Backend-Public: 10G
- Backend-Cluster: 10G
Just to add, while it's technically possible to mix these, it’s better to keep each type of traffic separate to avoid issues. If your VMs start generating heavy traffic, it could disrupt your storage connections, and that could lead to serious issues. Losing quorum could knock out one of your servers unexpectedly!
What outcome are you aiming for by connecting the hosts directly? Just curious!
I want to enhance fault tolerance and free up four ports.
Good to know! We have a dedicated management port as well. For our setup, it's just two small servers for a small office—likely won't add a third. We're using shared storage but it's DAS/SAS, so it's not over the network. I planned to use the 10G for VMkernel linking and the 1G as a failover connected to the switch.