catatonicprime

@draeath @sassdawe @codinghorror @jorge usually you have to step on the foot guns a few times anyway. During setup is a very good time to do that. The more the better. Because once it's in production, taking it out of production to fix the issues, can be hard.

I've been using kubernetes for about a year at home now. My build last year included:

• proxmox
• 6x vms running rke2
  • 3x system nodes
  • 3x agents
    • 3x these each have their own disk
• longhorn

So I'm all into the redundancy everything thing during the setup. So I opt for a raidz1 in zfs on proxmox for all the data drives. Then I add virtual disks to the agents. This was a very bad idea. And I ran it that way for 6mos. Because once it got running, we were stuck.

Similarly I've just installed a new switch, learned about VLANs, learned about LACP LAGGs, and have been re-iping the running cluster in place. It's been a challenge.

Tbf to the discussion though, I use a mixture of LLMs and forums when learning about these things. And the LLMs are definitely hit or miss once you get into the weeds. But people's communication style, particularly in the tech area, makes me hope for an asteroid strike.

I like kubernetes. I like longhorn. But God damn if I don't have a multi hour headache every time I want to apply os updates.