If you’re planning on running containers, tbh I would not. The required virtualization layer (due to macOS being BSD-based) is going to make containers WAY less efficient. You’re better off with a system you can easily slap proxmox (or whatever distro floats your boat) on.
Codeberg is pretty legit
Huh, really? Is there that much of a perf hit using passthrough? I’d have assumed that the bottleneck isn’t actually the PCIE, so much as it is the beefiness of the GPU crunching the model.
Bro this is a community for sharing knowledge and increasing the technical aptitude of fellow users by doing said sharing. Maybe instead of shitting on a pretty solid digest of the fundamentals of setting up something like this, try adding to the body of knowledge instead.
Wtf are you talking about. PCIe passthrough exists.
I’m in this picture and I don’t like it
Fair; I blame target fixation
Well sure, but the question was about gluetun, so I was trying to focus on that and the applications thereof. In terms of homelab stuff, I know a lot of people appreciate the containerized approach.
Oh yeah you can do it that way too, but if you want it all containerized, that’s roughly how to do it. That’s all I meant.
It’s convenient if you want to see gluetun up as the only way a container (say, your torrenting container) can get to the open net, in the interest of avoiding getting directly pinged by DMCA rats. That way, if the VPN goes down, your torrent client isn’t just downloading stuff nakedly. Also, if you want to set up different VPN connections for different containers, it’s pretty easy to set a handful of replica containers for that too.
lol this is so cursed, I love it.
Now include time, with tz offsets
Why do some people insist on doing big-endian dates? Little-endian (YYYY-MM-DD) is trivially alphanumerically sortable. Also, it’s part of the goddamn ISO standard for standardized date format.
Note that some issue devices have VT-x disabled and the bios locked down by Corp IT for one reason or another, so a VM may not actually be possible from the work issue device here.
Tbh, just run the Ubuntu box headless and ssh into it. You can do anything you’d need to. Even better, swap it to Debian or something like that, because Ubuntu is unfortunately kinda undergoing gradual enshitification lately.
Get out of my head charles
Haha, heisenbugs, always a fun time.
More seriously, I’d be surprised if this wasn’t a classic race condition
+1 - GL.iNet is cheap as hell and perfectly functional. It runs DD-WRT under the covers, and it’s super easy to “unlock” the full powers of the distro (literally a toggle in the basic ui, iirc). Used it on a longer trip my partner and I took a few months ago, and it was great! On-device storage is paltry, but, well, that’s not what it’s for.
deleted by creator
For SSDs, enterprise-tier drives typically set aside a lot more dedicated space for wear management. So in many cases, it’s actually a 1TB (or 2 -> 1.8T, etc) drive, but the disk controller/firmware only allows addressing a subset of the “true” capacity.
DAS enclosures are super easy to set up. Disk won’t be a big issue if you go that route.