I’ve now accomplished one of my previously-mentioned desires, namely, to use Packer to make VMs for Proxmox. After much battling with YAML and esoteric bash commands, I have succeeded in being able to spawn endless VMs, ready to go just how I like them. As with many projects I’ve done, this was thanks to someone else’s hard work; I merely customized it. The repo[s] are in two parts, and for my forks, only the Debian template has been customized....

A friend wanted to learn Linux, so I offered to spin up a VM under Proxmox. Done. Just kidding. I mean, that would work (assuming you handled port forwarding) if you were hitting an IP, but FQDNs are much easier for people to remember. Except ssh isn’t based on HTTP, so how do you forward them? One way is with nginx’s stream module. Something like this suffices. stream { upstream ssh { server $DEST_IP:$SSH_PORT; } server { listen $FORWARDED_PORT; proxy_pass ssh; } } And that’ll handle one person just fine....

This is not entirely an “it’s always DNS joke,” I promise. But for the record, it is always DNS. I have an (three, actually - two are boot drives/VM storage) NVMe drive in my server that serves multiple functions - a Plex cache, a ZFS Intent Log (I should get rid of that, as it’s shown to have 0% gains for me), and a general scratch disk for anytime I need a fast disk....

It’s coarse and rough and irritating and… it sucks. There’s a reason why cloud providers are wildly popular. Scaling in AWS is incredibly easy; not so much when you’re rolling your own. God help you if you’re doing this by cobbling together disparate groups of enterprise and consumer hardware. Oh wait, that’s exactly what I’m doing. I’ve had a Linux box since 2016 or so. I mean, I dual-booted every distro known to man with Windows when I was a teenager (including bootstrapped Gentoo, and I managed to get a crotchety HP printer to work with it), but I didn’t have a dedicated Linux box until then....

This is not so much of a SRE post, be ye warned, but it does discuss solving problems creatively. WARNING: Blindly following this will result in incurring costs from cloud providers, potentially quite a bit. It’s not my fault if you don’t calculate how much your task will cost. As an example, had I used AWS Comprehend for my sentiment analysis, I calculated the cost at ~$15,000. Why on earth would you need to know the gender of Github users, you ask?...