In my quest for my speed across my little lab, especially while I’m trying Longhorn v2 to my nvme drive in each host, I bought new to me NICs for each host. Running v1.7.1
I’m replacing x520-da2 with xxv710 dual port cards, both are Supermicro. Going from 10g on my Mikrotik sfp+ switch to the last 3 sfp28 ports on my Extreme 5420m switch. But know these are different families of drivers, so I’m expecting the naming to change which will probably disconnect the networks.
I do have kvm level access to the hosts so I can log in to the local terminal and edit files, just not sure where to look or how to reset the network. In XCP-ng I can log n locally and reset the management networking from the local text gui, but didn’t see anything like that for Harvester.
If you are going to be doing a lot of lab tests like this, don’t treat your lab nodes as pets that you are going to love forever. That very much is NOT what Harvester is. I suggest 3 nodes minimum ideally with a NFS share somewhere you can backup VM images to. Automate your Harvester and VM installs with PXE, terra form etc. You will learn so much more and be far less afraid of ripping and replacing/recovery and testing with fresh installs done in 5 minutes on your changes. If you need a loving home lab pet single node there are better options designed for this.I am not trying to be offensive here, I’m offering a different mindset.
Not offended because you are 100% correct. I do need to learn how to automate this stuff, and as a lab, I tear it apart regularly. The only thing I considering keeping is buying a license for MS Server 2025 for my AD, DNS, DHCP, etc. which runs on a standalone mini-pc. The longest I’ve had a version of Server Evaluation running has been about 18 months before trashing it and starting over, still not sure if I’ll keep it long enough to license.
That said, my learning curve is really steep on this, still trying to find my way around. I approached this from the wrong end, while Harvester covers up the need to understand Kubernetes for most tasks, you still need that baseline for more complex things. That’s a mistake I found the hard way, and something I will need to study up in the very near future.
As far as my lab goes, it’s an under powered cluster of three HP T740 thin clients. 64gb ram, 1tb sata, 1tb nvme, and the above mentioned networking with the addition of an Intel i226v in the WiFi slot, and the on board Realtek that I’m not planning to use. You’d be surprised how well this same cluster worked with vSphere8 back when I had a license, and how well it works with XCP-ng. I do have a small Truenas Scale on 10g that I use for shares, a little n100 powered board with sata drives.