NixOS uses QEMU extensively in automatically testing various configurations before each channel update. This means we automatically verify that the bootloader boots, EFI works, kernel operates correctly, and our entire install flow works. These tests run roughly every 4 hours. Because of these, even our nixos-unstable channel is very stable.
I highly recommend looking further at QEMU! FWIW though, it is dreadfully slow in an already virtualized environment. Any dedicated hardware with /dev/kvm should do the trick. I personally use Packet.net extensively for this task.
Here are the tests NixOS runs automatically: https://github.com/NixOS/nixpkgs/tree/master/nixos/tests
Here is a specific one, which I edited recently to verify that only users in the docker group could access the docker socket: https://github.com/NixOS/nixpkgs/blob/master/nixos/tests/docker.nix
At my day job, we’ve been looking at using QEMU to start up several MySQL daemons to test our replication topology orchestration toolkit.
This is unrelated to QEMU, but, wrt Packet.net, am I wrong in how I’m reading this? 36$/mo for a 4 core, 8GB RAM physical machine (the ‘tiny-but-mighty atom server’?). What’s the catch?
Nope, I don’t know of any catch. They’ve been incredible rock-solid stable. Their network is to die for, too, with their bonded 1G links. I’m a huge fan.