I feel like you understand the text book but didn’t know the application.

I’ve worked with Windows environments from 2003 until still today migrating to azure. The biggest skills gap with technicians and engineers administrating Windows is actually networking. This single point connects every single service server and user and yet dns, dhcp, routing and it’s protocols, link layer technologies like vlans interface configurations aggregation and more is so poorly understood that engineers and technicians often significantly mistake problems. Almost all issues happen around network layers 2-4 or layer 8 (the end user).

It doesn’t need to be first but no matter what os or component, networking is core and the single biggest return on investment for systems admin types.

Sure other basic skills are required but just being able to test TCP by telnet or understand each hop, and is the server listening? What process ID is listening? Did someone configure rdp off 3389 and that’s why it doesn’t work? Was the host file edited and that’s why it’s resolving some old ip for this hostname? Why is it going out the wan interface of the router when it should be going over an ipsec tunnel?

All this and more has nothing to do with Windows, and yet, anything that isn’t just user training or show and tell about how to do something, there’s a good chance it requires you to follow the networking layers to make sure behaviour is expected.

If dns resolved then it’s not blocked. You need to look at your network.

Bypass dns connect to the ip and port. What happens?

This won’t work, your wan ip isn’t dynamic, it’s on the ISP NAT network and your resulting ip to public services is shared across many customers. CG-NAT.

I don’t know where you work but don’t access your tailnet from a work device and ideally not their network.

Speaking to roku, you could buy a cheap raspberri pi and usb network port. One port to the network the other to roku. The pi can have a tailscale advertised network to the roku, and the roku probably needs nothing since everything is upstream including private tailscale 100.x.y.z networks which will be captured by your device in the middle raspberri pi.

I guess that’d cost like 40 ish dollars one time.

I went through a ghibli catalog watch while travelling in Japan one time, including Hiroshima peace Park (another “do once” thing. On the flight home I watched grave of the fireflies for the first time. I do not recommend watching it on a plane in public, when most everyone is sleeping so you try (fail) to keep your ugly sobbing to yourself.

Great movie

My kind of guy.

Australia just hopes the countries who handle the waste from the uranium we sell don’t you know make nuclear weapons with it. You know they’re good allies they wouldn’t do anything with that right? They certainly would never enrich the original…

Perhaps an anime with an ex-cop, ex-yakuza, a corgi, a young hacker, a gambling thief fugitive, and possibly the best intro and music in anime?

If Microsoft named it, it’s temporary. Let’s check intune no wait endpoint manager, no wait, intune again. First I’ll just make sure my login works in azure ad, no wait entra, no wait it’ll be copilot by the end of the year. Not to be confused with copilot (office) copilot (github) or copilot (azure) or power platform no wait copilot?

Would this not be the Linux subsystem on window? LSW

Which will be “If not Windows 11 or Mac os then report os string”. I don’t think they specifically took the time to research different OS’s and list them.

Wow I read through the blog post and though I’m not a developer I’ve compiled and built Linux packages and operating systems in the past so now I want to fly home and give your script a go myself.

I enjoyed your write up. I can’t comment on programming, but I enjoy a good journey and story.

My final takeaway is your image. I’ll keep it in mind. Interesting!

Make a YouTube on it and I’ll watch it. I’m not a coder though. But benchmarking and debunking is interesting. Either way it goes. Clear or complex the results come out it’ll be interesting.

What the heck is my Tuesday night doing posted here?

I’m far from an expert sorry, but my experience is so far so good (literally wizard configured in proxmox set and forget) even during a single disk lost. Performance for vm disks was great.

I can’t see why regular file would be any different.

I have 3 disks, one on each host, with ceph handling 2 copies (tolerant to 1 disk loss) distributed across them. That’s practically what I think you’re after.

I’m not sure about seeing the file system while all the hosts are all offline, but if you’ve got any one system with a valid copy online you should be able to see. I do. But my emphasis is generally get the host back online.

I’m not 100% sure what you’re trying to do but a mix of ceph as storage remote plus something like syncthing on a endpoint to send stuff to it might work? Syncthing might just work without ceph.

I also run zfs on an 8 disk nas that’s my primary storage with shares for my docker to send stuff, and media server to get it off. That’s just truenas scale. That way it handles data similarly. Zfs is also very good, but until scale came out, it wasn’t really possible to have the “add a compute node to expand your storage pool” which is how I want my vm hosts. Zfs scale looks way harder than ceph.

Not sure if any of that is helpful for your case but I recommend trying something if you’ve got spare hardware, and see how it goes on dummy data, then blow it away try something else. See how it acts when you take a machine offline. When you know what you want, do a final blow away and implement it with the way you learned to do it best.

3x Intel NUC 6th gen i5 (2 cores) 32gb RAM. Proxmox cluster with ceph.

I just ignored the limitation and tried with a single sodim of 32gb once (out of a laptop) and it worked fine, but just backed to 2x16gb dimms since the limit was still 2core of CPU. Lol.

Running that cluster 7 or so years now since I bought them new.

I suggest only running off shit tier since three nodes gives redundancy and enough performance. I’ve run entire proof of concepts for clients off them. Dual domain controllers and FC Rd gateway broker session hosts fxlogic etc. Back when Ms only just bought that tech. Meanwhile my home “ARR” just plugs on in docker containers. Even my opnsense router is virtual running on them. Just get a proper managed switch and take in the internet onto a vlan into the guest vm on a separate virtual NIC.

Point is, it’s still capable today.

I thought this was an onion article.

Sr-iov works already though? That’s not needed for this. The motherboard presents the pci bus to the guest regardless of what’s plugged in. Works fine.

This is when you want many guests to have shared graphics by partitioning a gpu. So the host still retains it and presents the graphics card to guests. You need to partition the ram up equally though, so useful only in VDI generally where you want a RTX A6000 like card to split to 10 guests each with 8gb of ram, and they share the gpu, but keep their individual video ram. Economy of scale can work out in graphics or maybe ML situations. Not so useful at home since you’ll probably have a Rtx 3080 with like 10-12gb of ram, and at most you wouldn’t want to split it below 8gb for modern games and partitions need to be equally sized. For 10g two = 2x5gb which would be a poor experience probably. Lots of frame stutters as it switches stuff between ram to video ram.

Hope that helps. Unless this technology unlocks better partitions it’s more about opening to vdi and machine learning in a full open source context like proxmox rather than just the driver being locked behind hyperv vmware and citrix hypervisor/xen and a big yearly license. Maybe it still needs that yearly license.

This is possible now, but in xen or vmware you need to buy a nvidia license to unlock this feature. You can trial it for a minute in a lab but you can’t give 4 guests each 2gb of vram on your graphics card without Nvidia specialist proprietary driver on both the host and the guest.

For vdi where you can buy 48gb rtx a6000 graphics cards, with architects (for example) each user getting each about 8gb each, you can 10 guests concurrently per card. Which at a few hundred architects scales better than buying many $5000 dollar workstations that struggle with WFH.

For a home user, maybe being able to split for your two kids on a standard rtx 3070 with what like 8gb might be OK? Probably not though.

Right now I have a hacky way that isn’t really supported in nvidia to split graphics cards to two guest vms but it’s neither license compatible or what I’d call “production ready”. I’d like proxmox to be able to handle this out of the box because it’s already in the kernel.

I’ve no idea what this means with licensing though. The yearly license cost to allow you to use your driver is actually stupidly expensive. The Rtx A series cards are already dumb money.

Either way it’s a good thing, but probably not much news for the average enthusiast