So what is the status of Thunderbolt networking under Linux? (Should I use it?)

So almost a year ago Wendel made a video about thunderbolt networking in the homelab. One of the big issues was that the kernel module was not well optimized and had high latency. Does anyone know if any work has been done? I am thinking about using it.

Right now I have a 3 node Proxmox cluster in my homelab connected together with a gigbit managed switch. Today I was doing some basic maintenance and it occurred to me that two of my machines have thunderbolt ports. I have been looking into some sort of high availability solution for a while and I am thinking about building a docker swarm. However, networking seemed to always be the issue as the switch could only go so fast. I do not want to run Ceph as it has to much overhead but I am thinking about either 2 node samba or nfs with fail over. I could have my non HA services like my vfio desktop and NAS running on the non thunderbolt device and then the docker swarm nodes on the thunderbolt devices with the link being used for the HA portion.

With this being said I am not sure what the state of thunderbolt networking is. High latency should not be much of an issue theoretically but I do not know if there is any other hiccups I need to worry about.

1 Like

until anyone else reports their experience, why not give it a go and let us know?

Wendell could himself have kept using it all this time, or just finished the project and moved on… but now it’s later on, perhaps stuff has improved?



I think I will give it ago for the hell of it


I could not get it working

Edit: it was the cable